qemu_process.c 181.8 KB
Newer Older
1
/*
2
 * qemu_process.c: QEMU process management
3
 *
4
 * Copyright (C) 2006-2015 Red Hat, Inc.
5 6 7 8 9 10 11 12 13 14 15 16
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
17
 * License along with this library.  If not, see
O
Osier Yang 已提交
18
 * <http://www.gnu.org/licenses/>.
19 20 21 22 23 24 25 26 27
 *
 */

#include <config.h>

#include <fcntl.h>
#include <unistd.h>
#include <signal.h>
#include <sys/stat.h>
R
Roman Bogorodskiy 已提交
28 29 30 31 32 33
#if defined(__linux__)
# include <linux/capability.h>
#elif defined(__FreeBSD__)
# include <sys/param.h>
# include <sys/cpuset.h>
#endif
34 35

#include "qemu_process.h"
36
#include "qemu_processpriv.h"
37 38 39 40 41 42 43
#include "qemu_domain.h"
#include "qemu_cgroup.h"
#include "qemu_capabilities.h"
#include "qemu_monitor.h"
#include "qemu_command.h"
#include "qemu_hostdev.h"
#include "qemu_hotplug.h"
44
#include "qemu_migration.h"
45
#include "qemu_interface.h"
46

47
#include "cpu/cpu.h"
48
#include "datatypes.h"
49
#include "virlog.h"
50
#include "virerror.h"
51
#include "viralloc.h"
52
#include "virhook.h"
E
Eric Blake 已提交
53
#include "virfile.h"
54
#include "virpidfile.h"
55 56
#include "c-ctype.h"
#include "nodeinfo.h"
57
#include "domain_audit.h"
58
#include "domain_nwfilter.h"
59
#include "locking/domain_lock.h"
60
#include "network/bridge_driver.h"
61
#include "viruuid.h"
62
#include "virprocess.h"
63
#include "virtime.h"
A
Ansis Atteka 已提交
64
#include "virnetdevtap.h"
65
#include "virnetdevopenvswitch.h"
66
#include "virnetdevmidonet.h"
67
#include "virbitmap.h"
68
#include "viratomic.h"
69
#include "virnuma.h"
70
#include "virstring.h"
71
#include "virhostdev.h"
72
#include "storage/storage_driver.h"
73
#include "configmake.h"
74
#include "nwfilter_conf.h"
75
#include "netdev_bandwidth_conf.h"
76 77 78

#define VIR_FROM_THIS VIR_FROM_QEMU

79 80
VIR_LOG_INIT("qemu.qemu_process");

81
#define ATTACH_POSTFIX ": attaching\n"
82 83 84
#define SHUTDOWN_POSTFIX ": shutting down\n"

/**
85
 * qemuProcessRemoveDomainStatus
86 87 88 89 90 91
 *
 * remove all state files of a domain from statedir
 *
 * Returns 0 on success
 */
static int
92
qemuProcessRemoveDomainStatus(virQEMUDriverPtr driver,
93 94 95 96
                              virDomainObjPtr vm)
{
    char ebuf[1024];
    char *file = NULL;
97
    qemuDomainObjPrivatePtr priv = vm->privateData;
98 99
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
    int ret = -1;
100

101
    if (virAsprintf(&file, "%s/%s.xml", cfg->stateDir, vm->def->name) < 0)
102
        goto cleanup;
103 104 105 106 107 108

    if (unlink(file) < 0 && errno != ENOENT && errno != ENOTDIR)
        VIR_WARN("Failed to remove domain XML for %s: %s",
                 vm->def->name, virStrerror(errno, ebuf, sizeof(ebuf)));
    VIR_FREE(file);

109 110 111
    if (priv->pidfile &&
        unlink(priv->pidfile) < 0 &&
        errno != ENOENT)
112 113 114
        VIR_WARN("Failed to remove PID file for %s: %s",
                 vm->def->name, virStrerror(errno, ebuf, sizeof(ebuf)));

115
    ret = 0;
116
 cleanup:
117 118
    virObjectUnref(cfg);
    return ret;
119 120 121 122
}


/* XXX figure out how to remove this */
123
extern virQEMUDriverPtr qemu_driver;
124

D
Daniel P. Berrange 已提交
125 126 127 128 129 130 131
/*
 * This is a callback registered with a qemuAgentPtr instance,
 * and to be invoked when the agent console hits an end of file
 * condition, or error, thus indicating VM shutdown should be
 * performed
 */
static void
132
qemuProcessHandleAgentEOF(qemuAgentPtr agent,
D
Daniel P. Berrange 已提交
133 134 135 136 137 138
                          virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv;

    VIR_DEBUG("Received EOF from agent on %p '%s'", vm, vm->def->name);

139
    virObjectLock(vm);
D
Daniel P. Berrange 已提交
140 141

    priv = vm->privateData;
142 143 144 145 146 147 148 149 150 151 152 153

    if (!priv->agent) {
        VIR_DEBUG("Agent freed already");
        goto unlock;
    }

    if (priv->beingDestroyed) {
        VIR_DEBUG("Domain is being destroyed, agent EOF is expected");
        goto unlock;
    }

    priv->agent = NULL;
D
Daniel P. Berrange 已提交
154

155
    virObjectUnlock(vm);
156 157

    qemuAgentClose(agent);
158 159
    return;

160
 unlock:
161 162
    virObjectUnlock(vm);
    return;
D
Daniel P. Berrange 已提交
163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179
}


/*
 * This is invoked when there is some kind of error
 * parsing data to/from the agent. The VM can continue
 * to run, but no further agent commands will be
 * allowed
 */
static void
qemuProcessHandleAgentError(qemuAgentPtr agent ATTRIBUTE_UNUSED,
                            virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv;

    VIR_DEBUG("Received error from agent on %p '%s'", vm, vm->def->name);

180
    virObjectLock(vm);
D
Daniel P. Berrange 已提交
181 182 183 184 185

    priv = vm->privateData;

    priv->agentError = true;

186
    virObjectUnlock(vm);
D
Daniel P. Berrange 已提交
187 188 189 190 191
}

static void qemuProcessHandleAgentDestroy(qemuAgentPtr agent,
                                          virDomainObjPtr vm)
{
192 193
    VIR_DEBUG("Received destroy agent=%p vm=%p", agent, vm);

194
    virObjectUnref(vm);
D
Daniel P. Berrange 已提交
195 196 197 198 199 200 201 202 203 204
}


static qemuAgentCallbacks agentCallbacks = {
    .destroy = qemuProcessHandleAgentDestroy,
    .eofNotify = qemuProcessHandleAgentEOF,
    .errorNotify = qemuProcessHandleAgentError,
};


205
int
206
qemuConnectAgent(virQEMUDriverPtr driver, virDomainObjPtr vm)
D
Daniel P. Berrange 已提交
207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret = -1;
    qemuAgentPtr agent = NULL;
    virDomainChrSourceDefPtr config = qemuFindAgentConfig(vm->def);

    if (!config)
        return 0;

    if (virSecurityManagerSetDaemonSocketLabel(driver->securityManager,
                                               vm->def) < 0) {
        VIR_ERROR(_("Failed to set security context for agent for %s"),
                  vm->def->name);
        goto cleanup;
    }

    /* Hold an extra reference because we can't allow 'vm' to be
     * deleted while the agent is active */
225
    virObjectRef(vm);
D
Daniel P. Berrange 已提交
226 227

    ignore_value(virTimeMillisNow(&priv->agentStart));
228
    virObjectUnlock(vm);
D
Daniel P. Berrange 已提交
229 230 231 232 233

    agent = qemuAgentOpen(vm,
                          config,
                          &agentCallbacks);

234
    virObjectLock(vm);
D
Daniel P. Berrange 已提交
235 236
    priv->agentStart = 0;

237 238 239 240 241 242 243 244 245 246 247
    if (agent == NULL)
        virObjectUnref(vm);

    if (!virDomainObjIsActive(vm)) {
        qemuAgentClose(agent);
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest crashed while connecting to the guest agent"));
        ret = -2;
        goto cleanup;
    }

D
Daniel P. Berrange 已提交
248 249 250 251
    if (virSecurityManagerClearSocketLabel(driver->securityManager,
                                           vm->def) < 0) {
        VIR_ERROR(_("Failed to clear security context for agent for %s"),
                  vm->def->name);
252
        qemuAgentClose(agent);
D
Daniel P. Berrange 已提交
253 254 255 256 257 258 259 260 261 262 263 264 265
        goto cleanup;
    }


    priv->agent = agent;

    if (priv->agent == NULL) {
        VIR_INFO("Failed to connect agent for %s", vm->def->name);
        goto cleanup;
    }

    ret = 0;

266
 cleanup:
D
Daniel P. Berrange 已提交
267 268 269 270
    return ret;
}


271
/*
272
 * This is a callback registered with a qemuMonitorPtr instance,
273 274 275 276 277 278
 * and to be invoked when the monitor console hits an end of file
 * condition, or error, thus indicating VM shutdown should be
 * performed
 */
static void
qemuProcessHandleMonitorEOF(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
279 280
                            virDomainObjPtr vm,
                            void *opaque)
281
{
282
    virQEMUDriverPtr driver = opaque;
283
    virObjectEventPtr event = NULL;
284
    qemuDomainObjPrivatePtr priv;
285 286 287
    int eventReason = VIR_DOMAIN_EVENT_STOPPED_SHUTDOWN;
    int stopReason = VIR_DOMAIN_SHUTOFF_SHUTDOWN;
    const char *auditReason = "shutdown";
288
    unsigned int stopFlags = 0;
289 290 291

    VIR_DEBUG("Received EOF on %p '%s'", vm, vm->def->name);

292
    virObjectLock(vm);
293

294 295 296 297 298 299 300
    priv = vm->privateData;

    if (priv->beingDestroyed) {
        VIR_DEBUG("Domain is being destroyed, EOF is expected");
        goto unlock;
    }

301 302
    if (!virDomainObjIsActive(vm)) {
        VIR_DEBUG("Domain %p is not active, ignoring EOF", vm);
303
        goto unlock;
304 305
    }

306
    if (priv->monJSON && !priv->gotShutdown) {
307 308
        VIR_DEBUG("Monitor connection to '%s' closed without SHUTDOWN event; "
                  "assuming the domain crashed", vm->def->name);
309
        eventReason = VIR_DOMAIN_EVENT_STOPPED_FAILED;
310
        stopReason = VIR_DOMAIN_SHUTOFF_CRASHED;
311
        auditReason = "failed";
312 313
    }

314 315
    if (priv->job.asyncJob == QEMU_ASYNC_JOB_MIGRATION_IN) {
        stopFlags |= VIR_QEMU_PROCESS_STOP_MIGRATED;
316 317
        qemuMigrationErrorSave(driver, vm->def->name,
                               qemuMonitorLastError(priv->mon));
318
    }
319

320
    event = virDomainEventLifecycleNewFromObj(vm,
321
                                     VIR_DOMAIN_EVENT_STOPPED,
322
                                     eventReason);
323
    qemuProcessStop(driver, vm, stopReason, stopFlags);
324
    virDomainAuditStop(vm, auditReason);
325

326
    if (!vm->persistent) {
327
        qemuDomainRemoveInactive(driver, vm);
328 329 330
        goto cleanup;
    }

331
 unlock:
332
    virObjectUnlock(vm);
333

334
 cleanup:
335
    if (event)
336
        qemuDomainEventQueue(driver, event);
337 338 339 340 341 342 343 344 345 346 347
}


/*
 * This is invoked when there is some kind of error
 * parsing data to/from the monitor. The VM can continue
 * to run, but no further monitor commands will be
 * allowed
 */
static void
qemuProcessHandleMonitorError(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
348 349
                              virDomainObjPtr vm,
                              void *opaque)
350
{
351
    virQEMUDriverPtr driver = opaque;
352
    virObjectEventPtr event = NULL;
353 354 355

    VIR_DEBUG("Received error on %p '%s'", vm, vm->def->name);

356
    virObjectLock(vm);
357

358
    ((qemuDomainObjPrivatePtr) vm->privateData)->monError = true;
359 360 361 362
    event = virDomainEventControlErrorNewFromObj(vm);
    if (event)
        qemuDomainEventQueue(driver, event);

363
    virObjectUnlock(vm);
364 365 366
}


367
virDomainDiskDefPtr
368 369 370
qemuProcessFindDomainDiskByAlias(virDomainObjPtr vm,
                                 const char *alias)
{
371
    size_t i;
372 373 374 375 376 377 378 379 380 381 382 383

    if (STRPREFIX(alias, QEMU_DRIVE_HOST_PREFIX))
        alias += strlen(QEMU_DRIVE_HOST_PREFIX);

    for (i = 0; i < vm->def->ndisks; i++) {
        virDomainDiskDefPtr disk;

        disk = vm->def->disks[i];
        if (disk->info.alias != NULL && STREQ(disk->info.alias, alias))
            return disk;
    }

384 385 386
    virReportError(VIR_ERR_INTERNAL_ERROR,
                   _("no disk found with alias %s"),
                   alias);
387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402
    return NULL;
}

static int
qemuProcessGetVolumeQcowPassphrase(virConnectPtr conn,
                                   virDomainDiskDefPtr disk,
                                   char **secretRet,
                                   size_t *secretLen)
{
    virSecretPtr secret;
    char *passphrase;
    unsigned char *data;
    size_t size;
    int ret = -1;
    virStorageEncryptionPtr enc;

403
    if (!disk->src->encryption) {
404 405
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("disk %s does not have any encryption information"),
406
                       disk->src->path);
407 408
        return -1;
    }
409
    enc = disk->src->encryption;
410 411

    if (!conn) {
412 413
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       "%s", _("cannot find secrets without a connection"));
414 415 416 417
        goto cleanup;
    }

    if (conn->secretDriver == NULL ||
418 419
        conn->secretDriver->secretLookupByUUID == NULL ||
        conn->secretDriver->secretGetValue == NULL) {
420 421
        virReportError(VIR_ERR_OPERATION_INVALID, "%s",
                       _("secret storage not supported"));
422 423 424 425 426 427 428
        goto cleanup;
    }

    if (enc->format != VIR_STORAGE_ENCRYPTION_FORMAT_QCOW ||
        enc->nsecrets != 1 ||
        enc->secrets[0]->type !=
        VIR_STORAGE_ENCRYPTION_SECRET_TYPE_PASSPHRASE) {
429
        virReportError(VIR_ERR_XML_ERROR,
430 431
                       _("invalid <encryption> for volume %s"),
                       virDomainDiskGetSource(disk));
432 433 434
        goto cleanup;
    }

435 436
    secret = conn->secretDriver->secretLookupByUUID(conn,
                                                    enc->secrets[0]->uuid);
437 438
    if (secret == NULL)
        goto cleanup;
439 440
    data = conn->secretDriver->secretGetValue(secret, &size, 0,
                                              VIR_SECRET_GET_VALUE_INTERNAL_CALL);
441
    virObjectUnref(secret);
442 443 444 445 446 447
    if (data == NULL)
        goto cleanup;

    if (memchr(data, '\0', size) != NULL) {
        memset(data, 0, size);
        VIR_FREE(data);
448 449
        virReportError(VIR_ERR_XML_ERROR,
                       _("format='qcow' passphrase for %s must not contain a "
450
                         "'\\0'"), virDomainDiskGetSource(disk));
451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469
        goto cleanup;
    }

    if (VIR_ALLOC_N(passphrase, size + 1) < 0) {
        memset(data, 0, size);
        VIR_FREE(data);
        goto cleanup;
    }
    memcpy(passphrase, data, size);
    passphrase[size] = '\0';

    memset(data, 0, size);
    VIR_FREE(data);

    *secretRet = passphrase;
    *secretLen = size;

    ret = 0;

470
 cleanup:
471 472 473 474 475 476 477 478 479
    return ret;
}

static int
qemuProcessFindVolumeQcowPassphrase(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                                    virConnectPtr conn,
                                    virDomainObjPtr vm,
                                    const char *path,
                                    char **secretRet,
480 481
                                    size_t *secretLen,
                                    void *opaque ATTRIBUTE_UNUSED)
482 483 484 485
{
    virDomainDiskDefPtr disk;
    int ret = -1;

486
    virObjectLock(vm);
487 488 489 490
    if (!(disk = virDomainDiskByName(vm->def, path, true))) {
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("no disk found with path %s"),
                       path);
491
        goto cleanup;
492
    }
493 494 495

    ret = qemuProcessGetVolumeQcowPassphrase(conn, disk, secretRet, secretLen);

496
 cleanup:
497
    virObjectUnlock(vm);
498 499 500 501 502 503
    return ret;
}


static int
qemuProcessHandleReset(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
504 505
                       virDomainObjPtr vm,
                       void *opaque)
506
{
507
    virQEMUDriverPtr driver = opaque;
508
    virObjectEventPtr event;
509
    qemuDomainObjPrivatePtr priv;
510
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
511

512
    virObjectLock(vm);
513

514
    event = virDomainEventRebootNewFromObj(vm);
515 516 517
    priv = vm->privateData;
    if (priv->agent)
        qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_RESET);
518 519 520

    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
        VIR_WARN("Failed to save status on vm %s", vm->def->name);
521

522
    virObjectUnlock(vm);
523

524
    if (event)
525 526
        qemuDomainEventQueue(driver, event);

527
    virObjectUnref(cfg);
528 529 530 531
    return 0;
}


532 533 534 535 536 537 538 539 540 541 542
/*
 * Since we have the '-no-shutdown' flag set, the
 * QEMU process will currently have guest OS shutdown
 * and the CPUS stopped. To fake the reboot, we thus
 * want todo a reset of the virtual hardware, followed
 * by restart of the CPUs. This should result in the
 * guest OS booting up again
 */
static void
qemuProcessFakeReboot(void *opaque)
{
543
    virQEMUDriverPtr driver = qemu_driver;
544 545
    virDomainObjPtr vm = opaque;
    qemuDomainObjPrivatePtr priv = vm->privateData;
546
    virObjectEventPtr event = NULL;
547
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
548
    virDomainRunningReason reason = VIR_DOMAIN_RUNNING_BOOTED;
549
    int ret = -1, rc;
550

551
    VIR_DEBUG("vm=%p", vm);
552
    virObjectLock(vm);
553
    if (qemuDomainObjBeginJob(driver, vm, QEMU_JOB_MODIFY) < 0)
554 555 556
        goto cleanup;

    if (!virDomainObjIsActive(vm)) {
557 558
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("guest unexpectedly quit"));
559 560 561
        goto endjob;
    }

562
    qemuDomainObjEnterMonitor(driver, vm);
563 564 565
    rc = qemuMonitorSystemReset(priv->mon);

    if (qemuDomainObjExitMonitor(driver, vm) < 0)
566 567
        goto endjob;

568
    if (rc < 0)
569 570
        goto endjob;

571 572 573
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_CRASHED)
        reason = VIR_DOMAIN_RUNNING_CRASHED;

574
    if (qemuProcessStartCPUs(driver, vm, NULL,
575
                             reason,
576
                             QEMU_ASYNC_JOB_NONE) < 0) {
577
        if (virGetLastError() == NULL)
578 579
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           "%s", _("resume operation failed"));
580 581
        goto endjob;
    }
582
    priv->gotShutdown = false;
583
    event = virDomainEventLifecycleNewFromObj(vm,
584 585 586
                                     VIR_DOMAIN_EVENT_RESUMED,
                                     VIR_DOMAIN_EVENT_RESUMED_UNPAUSED);

587 588 589 590 591
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
        VIR_WARN("Unable to save status on vm %s after state change",
                 vm->def->name);
    }

592 593
    ret = 0;

594
 endjob:
595
    qemuDomainObjEndJob(driver, vm);
596

597
 cleanup:
598 599
    if (ret == -1)
        ignore_value(qemuProcessKill(vm, VIR_QEMU_PROCESS_KILL_FORCE));
M
Michal Privoznik 已提交
600
    virDomainObjEndAPI(&vm);
601 602
    if (event)
        qemuDomainEventQueue(driver, event);
603
    virObjectUnref(cfg);
604 605 606
}


607
void
608
qemuProcessShutdownOrReboot(virQEMUDriverPtr driver,
609
                            virDomainObjPtr vm)
610
{
611 612 613
    qemuDomainObjPrivatePtr priv = vm->privateData;

    if (priv->fakeReboot) {
614
        qemuDomainSetFakeReboot(driver, vm, false);
615
        virObjectRef(vm);
616 617 618 619 620
        virThread th;
        if (virThreadCreate(&th,
                            false,
                            qemuProcessFakeReboot,
                            vm) < 0) {
621
            VIR_ERROR(_("Failed to create reboot thread, killing domain"));
622
            ignore_value(qemuProcessKill(vm, VIR_QEMU_PROCESS_KILL_NOWAIT));
623
            virObjectUnref(vm);
624 625
        }
    } else {
626
        ignore_value(qemuProcessKill(vm, VIR_QEMU_PROCESS_KILL_NOWAIT));
627
    }
628
}
629

630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657

static int
qemuProcessHandleEvent(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                       virDomainObjPtr vm,
                       const char *eventName,
                       long long seconds,
                       unsigned int micros,
                       const char *details,
                       void *opaque)
{
    virQEMUDriverPtr driver = opaque;
    virObjectEventPtr event = NULL;

    VIR_DEBUG("vm=%p", vm);

    virObjectLock(vm);
    event = virDomainQemuMonitorEventNew(vm->def->id, vm->def->name,
                                         vm->def->uuid, eventName,
                                         seconds, micros, details);

    virObjectUnlock(vm);
    if (event)
        qemuDomainEventQueue(driver, event);

    return 0;
}


658 659
static int
qemuProcessHandleShutdown(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
660 661
                          virDomainObjPtr vm,
                          void *opaque)
662
{
663
    virQEMUDriverPtr driver = opaque;
664
    qemuDomainObjPrivatePtr priv;
665
    virObjectEventPtr event = NULL;
666
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
667

668 669
    VIR_DEBUG("vm=%p", vm);

670
    virObjectLock(vm);
671 672 673 674 675 676

    priv = vm->privateData;
    if (priv->gotShutdown) {
        VIR_DEBUG("Ignoring repeated SHUTDOWN event from domain %s",
                  vm->def->name);
        goto unlock;
677 678 679 680
    } else if (!virDomainObjIsActive(vm)) {
        VIR_DEBUG("Ignoring SHUTDOWN event from inactive domain %s",
                  vm->def->name);
        goto unlock;
681 682 683 684 685 686 687 688
    }
    priv->gotShutdown = true;

    VIR_DEBUG("Transitioned guest %s to shutdown state",
              vm->def->name);
    virDomainObjSetState(vm,
                         VIR_DOMAIN_SHUTDOWN,
                         VIR_DOMAIN_SHUTDOWN_UNKNOWN);
689
    event = virDomainEventLifecycleNewFromObj(vm,
690 691 692
                                     VIR_DOMAIN_EVENT_SHUTDOWN,
                                     VIR_DOMAIN_EVENT_SHUTDOWN_FINISHED);

693
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
694 695 696 697
        VIR_WARN("Unable to save status on vm %s after state change",
                 vm->def->name);
    }

698 699 700
    if (priv->agent)
        qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_SHUTDOWN);

701 702
    qemuProcessShutdownOrReboot(driver, vm);

703
 unlock:
704
    virObjectUnlock(vm);
705
    if (event)
706
        qemuDomainEventQueue(driver, event);
707
    virObjectUnref(cfg);
708

709 710 711 712 713 714
    return 0;
}


static int
qemuProcessHandleStop(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
715 716
                      virDomainObjPtr vm,
                      void *opaque)
717
{
718
    virQEMUDriverPtr driver = opaque;
719
    virObjectEventPtr event = NULL;
720
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
721

722
    virObjectLock(vm);
J
Jiri Denemark 已提交
723
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
724
        qemuDomainObjPrivatePtr priv = vm->privateData;
725

726
        if (priv->gotShutdown) {
727 728
            VIR_DEBUG("Ignoring STOP event after SHUTDOWN");
            goto unlock;
729 730
        }

731 732
        VIR_DEBUG("Transitioned guest %s to paused state",
                  vm->def->name);
733

734 735 736
        if (priv->job.current)
            ignore_value(virTimeMillisNow(&priv->job.current->stopped));

737
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_UNKNOWN);
738
        event = virDomainEventLifecycleNewFromObj(vm,
739 740 741
                                         VIR_DOMAIN_EVENT_SUSPENDED,
                                         VIR_DOMAIN_EVENT_SUSPENDED_PAUSED);

742 743 744 745 746
        VIR_FREE(priv->lockState);
        if (virDomainLockProcessPause(driver->lockManager, vm, &priv->lockState) < 0)
            VIR_WARN("Unable to release lease on %s", vm->def->name);
        VIR_DEBUG("Preserving lock state '%s'", NULLSTR(priv->lockState));

747
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
748 749 750
            VIR_WARN("Unable to save status on vm %s after state change",
                     vm->def->name);
        }
751
    }
752

753
 unlock:
754
    virObjectUnlock(vm);
755
    if (event)
756
        qemuDomainEventQueue(driver, event);
757
    virObjectUnref(cfg);
758 759 760 761 762

    return 0;
}


763 764
static int
qemuProcessHandleResume(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
765 766
                        virDomainObjPtr vm,
                        void *opaque)
767
{
768
    virQEMUDriverPtr driver = opaque;
769
    virObjectEventPtr event = NULL;
770
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
771

772
    virObjectLock(vm);
773 774 775 776 777 778 779 780 781 782 783 784 785
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_PAUSED) {
        qemuDomainObjPrivatePtr priv = vm->privateData;

        if (priv->gotShutdown) {
            VIR_DEBUG("Ignoring RESUME event after SHUTDOWN");
            goto unlock;
        }

        VIR_DEBUG("Transitioned guest %s out of paused into resumed state",
                  vm->def->name);

        virDomainObjSetState(vm, VIR_DOMAIN_RUNNING,
                                 VIR_DOMAIN_RUNNING_UNPAUSED);
786
        event = virDomainEventLifecycleNewFromObj(vm,
787 788 789 790
                                         VIR_DOMAIN_EVENT_RESUMED,
                                         VIR_DOMAIN_EVENT_RESUMED_UNPAUSED);

        VIR_DEBUG("Using lock state '%s' on resume event", NULLSTR(priv->lockState));
791
        if (virDomainLockProcessResume(driver->lockManager, cfg->uri,
792 793 794 795 796 797 798 799 800
                                       vm, priv->lockState) < 0) {
            /* Don't free priv->lockState on error, because we need
             * to make sure we have state still present if the user
             * tries to resume again
             */
            goto unlock;
        }
        VIR_FREE(priv->lockState);

801
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
802 803 804 805 806
            VIR_WARN("Unable to save status on vm %s after state change",
                     vm->def->name);
        }
    }

807
 unlock:
808
    virObjectUnlock(vm);
809
    if (event)
810
        qemuDomainEventQueue(driver, event);
811
    virObjectUnref(cfg);
812 813 814
    return 0;
}

815 816 817
static int
qemuProcessHandleRTCChange(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                           virDomainObjPtr vm,
818 819
                           long long offset,
                           void *opaque)
820
{
821
    virQEMUDriverPtr driver = opaque;
822
    virObjectEventPtr event = NULL;
823
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
824

825
    virObjectLock(vm);
826

827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843
    if (vm->def->clock.offset == VIR_DOMAIN_CLOCK_OFFSET_VARIABLE) {
        /* when a basedate is manually given on the qemu commandline
         * rather than simply "-rtc base=utc", the offset sent by qemu
         * in this event is *not* the new offset from UTC, but is
         * instead the new offset from the *original basedate* +
         * uptime. For example, if the original offset was 3600 and
         * the guest clock has been advanced by 10 seconds, qemu will
         * send "10" in the event - this means that the new offset
         * from UTC is 3610, *not* 10. If the guest clock is advanced
         * by another 10 seconds, qemu will now send "20" - i.e. each
         * event is the sum of the most recent change and all previous
         * changes since the domain was started. Fortunately, we have
         * saved the initial offset in "adjustment0", so to arrive at
         * the proper new "adjustment", we just add the most recent
         * offset to adjustment0.
         */
        offset += vm->def->clock.data.variable.adjustment0;
844
        vm->def->clock.data.variable.adjustment = offset;
845

846 847 848 849 850
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
           VIR_WARN("unable to save domain status with RTC change");
    }

    event = virDomainEventRTCChangeNewFromObj(vm, offset);
851

852
    virObjectUnlock(vm);
853

854
    if (event)
855
        qemuDomainEventQueue(driver, event);
856
    virObjectUnref(cfg);
857 858 859 860 861 862 863
    return 0;
}


static int
qemuProcessHandleWatchdog(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                          virDomainObjPtr vm,
864 865
                          int action,
                          void *opaque)
866
{
867
    virQEMUDriverPtr driver = opaque;
868 869
    virObjectEventPtr watchdogEvent = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
870
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
871

872
    virObjectLock(vm);
873 874 875
    watchdogEvent = virDomainEventWatchdogNewFromObj(vm, action);

    if (action == VIR_DOMAIN_EVENT_WATCHDOG_PAUSE &&
J
Jiri Denemark 已提交
876
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
877
        qemuDomainObjPrivatePtr priv = vm->privateData;
878 879
        VIR_DEBUG("Transitioned guest %s to paused state due to watchdog", vm->def->name);

J
Jiri Denemark 已提交
880
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_WATCHDOG);
881
        lifecycleEvent = virDomainEventLifecycleNewFromObj(vm,
882 883 884
                                                  VIR_DOMAIN_EVENT_SUSPENDED,
                                                  VIR_DOMAIN_EVENT_SUSPENDED_WATCHDOG);

885 886 887 888 889
        VIR_FREE(priv->lockState);
        if (virDomainLockProcessPause(driver->lockManager, vm, &priv->lockState) < 0)
            VIR_WARN("Unable to release lease on %s", vm->def->name);
        VIR_DEBUG("Preserving lock state '%s'", NULLSTR(priv->lockState));

890
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
891 892 893
            VIR_WARN("Unable to save status on vm %s after watchdog event",
                     vm->def->name);
        }
894 895 896
    }

    if (vm->def->watchdog->action == VIR_DOMAIN_WATCHDOG_ACTION_DUMP) {
C
Chen Fan 已提交
897 898 899 900 901
        struct qemuProcessEvent *processEvent;
        if (VIR_ALLOC(processEvent) == 0) {
            processEvent->eventType = QEMU_PROCESS_EVENT_WATCHDOG;
            processEvent->action = VIR_DOMAIN_WATCHDOG_ACTION_DUMP;
            processEvent->vm = vm;
W
Wen Congyang 已提交
902 903 904
            /* Hold an extra reference because we can't allow 'vm' to be
             * deleted before handling watchdog event is finished.
             */
905
            virObjectRef(vm);
C
Chen Fan 已提交
906
            if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
907
                if (!virObjectUnref(vm))
908
                    vm = NULL;
C
Chen Fan 已提交
909
                VIR_FREE(processEvent);
W
Wen Congyang 已提交
910
            }
911
        }
912 913
    }

914
    if (vm)
915
        virObjectUnlock(vm);
916 917 918 919
    if (watchdogEvent)
        qemuDomainEventQueue(driver, watchdogEvent);
    if (lifecycleEvent)
        qemuDomainEventQueue(driver, lifecycleEvent);
920

921
    virObjectUnref(cfg);
922 923 924 925 926 927 928 929 930
    return 0;
}


static int
qemuProcessHandleIOError(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                         virDomainObjPtr vm,
                         const char *diskAlias,
                         int action,
931 932
                         const char *reason,
                         void *opaque)
933
{
934
    virQEMUDriverPtr driver = opaque;
935 936 937
    virObjectEventPtr ioErrorEvent = NULL;
    virObjectEventPtr ioErrorEvent2 = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
938 939 940
    const char *srcPath;
    const char *devAlias;
    virDomainDiskDefPtr disk;
941
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
942

943
    virObjectLock(vm);
944 945 946
    disk = qemuProcessFindDomainDiskByAlias(vm, diskAlias);

    if (disk) {
947
        srcPath = virDomainDiskGetSource(disk);
948 949 950 951 952 953 954 955 956 957
        devAlias = disk->info.alias;
    } else {
        srcPath = "";
        devAlias = "";
    }

    ioErrorEvent = virDomainEventIOErrorNewFromObj(vm, srcPath, devAlias, action);
    ioErrorEvent2 = virDomainEventIOErrorReasonNewFromObj(vm, srcPath, devAlias, action, reason);

    if (action == VIR_DOMAIN_EVENT_IO_ERROR_PAUSE &&
J
Jiri Denemark 已提交
958
        virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
959
        qemuDomainObjPrivatePtr priv = vm->privateData;
960 961
        VIR_DEBUG("Transitioned guest %s to paused state due to IO error", vm->def->name);

962 963 964
        if (priv->signalIOError)
            virDomainObjBroadcast(vm);

J
Jiri Denemark 已提交
965
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_IOERROR);
966
        lifecycleEvent = virDomainEventLifecycleNewFromObj(vm,
967 968 969
                                                  VIR_DOMAIN_EVENT_SUSPENDED,
                                                  VIR_DOMAIN_EVENT_SUSPENDED_IOERROR);

970 971 972 973 974
        VIR_FREE(priv->lockState);
        if (virDomainLockProcessPause(driver->lockManager, vm, &priv->lockState) < 0)
            VIR_WARN("Unable to release lease on %s", vm->def->name);
        VIR_DEBUG("Preserving lock state '%s'", NULLSTR(priv->lockState));

975
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
976 977
            VIR_WARN("Unable to save status on vm %s after IO error", vm->def->name);
    }
978
    virObjectUnlock(vm);
979

980 981 982 983 984 985
    if (ioErrorEvent)
        qemuDomainEventQueue(driver, ioErrorEvent);
    if (ioErrorEvent2)
        qemuDomainEventQueue(driver, ioErrorEvent2);
    if (lifecycleEvent)
        qemuDomainEventQueue(driver, lifecycleEvent);
986
    virObjectUnref(cfg);
987 988 989
    return 0;
}

990 991 992 993 994
static int
qemuProcessHandleBlockJob(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                          virDomainObjPtr vm,
                          const char *diskAlias,
                          int type,
995 996
                          int status,
                          void *opaque)
997
{
998
    virQEMUDriverPtr driver = opaque;
999
    struct qemuProcessEvent *processEvent = NULL;
1000
    virDomainDiskDefPtr disk;
1001
    qemuDomainDiskPrivatePtr diskPriv;
1002
    char *data = NULL;
1003

1004
    virObjectLock(vm);
1005

1006 1007
    VIR_DEBUG("Block job for device %s (domain: %p,%s) type %d status %d",
              diskAlias, vm, vm->def->name, type, status);
1008

1009
    if (!(disk = qemuProcessFindDomainDiskByAlias(vm, diskAlias)))
1010
        goto error;
1011
    diskPriv = QEMU_DOMAIN_DISK_PRIVATE(disk);
1012

1013
    if (diskPriv->blockJobSync) {
1014
        /* We have a SYNC API waiting for this event, dispatch it back */
1015 1016
        diskPriv->blockJobType = type;
        diskPriv->blockJobStatus = status;
1017
        virDomainObjBroadcast(vm);
1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029
    } else {
        /* there is no waiting SYNC API, dispatch the update to a thread */
        if (VIR_ALLOC(processEvent) < 0)
            goto error;

        processEvent->eventType = QEMU_PROCESS_EVENT_BLOCK_JOB;
        if (VIR_STRDUP(data, diskAlias) < 0)
            goto error;
        processEvent->data = data;
        processEvent->vm = vm;
        processEvent->action = type;
        processEvent->status = status;
1030

1031 1032 1033 1034 1035
        virObjectRef(vm);
        if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
            ignore_value(virObjectUnref(vm));
            goto error;
        }
1036 1037
    }

1038
 cleanup:
1039
    virObjectUnlock(vm);
1040
    return 0;
1041 1042 1043 1044 1045
 error:
    if (processEvent)
        VIR_FREE(processEvent->data);
    VIR_FREE(processEvent);
    goto cleanup;
1046
}
1047

1048

1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060
static int
qemuProcessHandleGraphics(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                          virDomainObjPtr vm,
                          int phase,
                          int localFamily,
                          const char *localNode,
                          const char *localService,
                          int remoteFamily,
                          const char *remoteNode,
                          const char *remoteService,
                          const char *authScheme,
                          const char *x509dname,
1061 1062
                          const char *saslUsername,
                          void *opaque)
1063
{
1064
    virQEMUDriverPtr driver = opaque;
1065
    virObjectEventPtr event;
1066 1067 1068
    virDomainEventGraphicsAddressPtr localAddr = NULL;
    virDomainEventGraphicsAddressPtr remoteAddr = NULL;
    virDomainEventGraphicsSubjectPtr subject = NULL;
1069
    size_t i;
1070 1071

    if (VIR_ALLOC(localAddr) < 0)
1072
        goto error;
1073
    localAddr->family = localFamily;
1074 1075 1076
    if (VIR_STRDUP(localAddr->service, localService) < 0 ||
        VIR_STRDUP(localAddr->node, localNode) < 0)
        goto error;
1077 1078

    if (VIR_ALLOC(remoteAddr) < 0)
1079
        goto error;
1080
    remoteAddr->family = remoteFamily;
1081 1082 1083
    if (VIR_STRDUP(remoteAddr->service, remoteService) < 0 ||
        VIR_STRDUP(remoteAddr->node, remoteNode) < 0)
        goto error;
1084 1085

    if (VIR_ALLOC(subject) < 0)
1086
        goto error;
1087 1088
    if (x509dname) {
        if (VIR_REALLOC_N(subject->identities, subject->nidentity+1) < 0)
1089
            goto error;
1090
        subject->nidentity++;
1091 1092 1093
        if (VIR_STRDUP(subject->identities[subject->nidentity-1].type, "x509dname") < 0 ||
            VIR_STRDUP(subject->identities[subject->nidentity-1].name, x509dname) < 0)
            goto error;
1094 1095 1096
    }
    if (saslUsername) {
        if (VIR_REALLOC_N(subject->identities, subject->nidentity+1) < 0)
1097
            goto error;
1098
        subject->nidentity++;
1099 1100 1101
        if (VIR_STRDUP(subject->identities[subject->nidentity-1].type, "saslUsername") < 0 ||
            VIR_STRDUP(subject->identities[subject->nidentity-1].name, saslUsername) < 0)
            goto error;
1102 1103
    }

1104
    virObjectLock(vm);
1105
    event = virDomainEventGraphicsNewFromObj(vm, phase, localAddr, remoteAddr, authScheme, subject);
1106
    virObjectUnlock(vm);
1107

1108
    if (event)
1109 1110 1111 1112
        qemuDomainEventQueue(driver, event);

    return 0;

1113
 error:
1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124
    if (localAddr) {
        VIR_FREE(localAddr->service);
        VIR_FREE(localAddr->node);
        VIR_FREE(localAddr);
    }
    if (remoteAddr) {
        VIR_FREE(remoteAddr->service);
        VIR_FREE(remoteAddr->node);
        VIR_FREE(remoteAddr);
    }
    if (subject) {
1125
        for (i = 0; i < subject->nidentity; i++) {
1126 1127 1128 1129 1130 1131 1132 1133 1134 1135
            VIR_FREE(subject->identities[i].type);
            VIR_FREE(subject->identities[i].name);
        }
        VIR_FREE(subject->identities);
        VIR_FREE(subject);
    }

    return -1;
}

1136 1137 1138 1139
static int
qemuProcessHandleTrayChange(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                            virDomainObjPtr vm,
                            const char *devAlias,
1140 1141
                            int reason,
                            void *opaque)
1142
{
1143
    virQEMUDriverPtr driver = opaque;
1144
    virObjectEventPtr event = NULL;
1145
    virDomainDiskDefPtr disk;
1146
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
1147

1148
    virObjectLock(vm);
1149 1150 1151 1152 1153 1154
    disk = qemuProcessFindDomainDiskByAlias(vm, devAlias);

    if (disk) {
        event = virDomainEventTrayChangeNewFromObj(vm,
                                                   devAlias,
                                                   reason);
1155 1156 1157 1158 1159 1160
        /* Update disk tray status */
        if (reason == VIR_DOMAIN_EVENT_TRAY_CHANGE_OPEN)
            disk->tray_status = VIR_DOMAIN_DISK_TRAY_OPEN;
        else if (reason == VIR_DOMAIN_EVENT_TRAY_CHANGE_CLOSE)
            disk->tray_status = VIR_DOMAIN_DISK_TRAY_CLOSED;

1161
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
1162 1163 1164
            VIR_WARN("Unable to save status on vm %s after tray moved event",
                     vm->def->name);
        }
1165 1166

        virDomainObjBroadcast(vm);
1167 1168
    }

1169
    virObjectUnlock(vm);
1170
    if (event)
1171
        qemuDomainEventQueue(driver, event);
1172
    virObjectUnref(cfg);
1173 1174 1175
    return 0;
}

O
Osier Yang 已提交
1176 1177
static int
qemuProcessHandlePMWakeup(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
1178 1179
                          virDomainObjPtr vm,
                          void *opaque)
O
Osier Yang 已提交
1180
{
1181
    virQEMUDriverPtr driver = opaque;
1182 1183
    virObjectEventPtr event = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
1184
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
O
Osier Yang 已提交
1185

1186
    virObjectLock(vm);
O
Osier Yang 已提交
1187 1188
    event = virDomainEventPMWakeupNewFromObj(vm);

1189 1190 1191 1192 1193 1194 1195 1196 1197
    /* Don't set domain status back to running if it wasn't paused
     * from guest side, otherwise it can just cause confusion.
     */
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_PMSUSPENDED) {
        VIR_DEBUG("Transitioned guest %s from pmsuspended to running "
                  "state due to QMP wakeup event", vm->def->name);

        virDomainObjSetState(vm, VIR_DOMAIN_RUNNING,
                             VIR_DOMAIN_RUNNING_WAKEUP);
1198
        lifecycleEvent = virDomainEventLifecycleNewFromObj(vm,
1199 1200 1201
                                                  VIR_DOMAIN_EVENT_STARTED,
                                                  VIR_DOMAIN_EVENT_STARTED_WAKEUP);

1202
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
1203 1204 1205 1206 1207
            VIR_WARN("Unable to save status on vm %s after wakeup event",
                     vm->def->name);
        }
    }

1208
    virObjectUnlock(vm);
1209 1210 1211 1212
    if (event)
        qemuDomainEventQueue(driver, event);
    if (lifecycleEvent)
        qemuDomainEventQueue(driver, lifecycleEvent);
1213
    virObjectUnref(cfg);
O
Osier Yang 已提交
1214 1215
    return 0;
}
1216

O
Osier Yang 已提交
1217 1218
static int
qemuProcessHandlePMSuspend(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
1219 1220
                           virDomainObjPtr vm,
                           void *opaque)
O
Osier Yang 已提交
1221
{
1222
    virQEMUDriverPtr driver = opaque;
1223 1224
    virObjectEventPtr event = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
1225
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
O
Osier Yang 已提交
1226

1227
    virObjectLock(vm);
O
Osier Yang 已提交
1228 1229
    event = virDomainEventPMSuspendNewFromObj(vm);

1230
    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
1231
        qemuDomainObjPrivatePtr priv = vm->privateData;
1232 1233 1234 1235 1236
        VIR_DEBUG("Transitioned guest %s to pmsuspended state due to "
                  "QMP suspend event", vm->def->name);

        virDomainObjSetState(vm, VIR_DOMAIN_PMSUSPENDED,
                             VIR_DOMAIN_PMSUSPENDED_UNKNOWN);
J
Jiri Denemark 已提交
1237
        lifecycleEvent =
1238
            virDomainEventLifecycleNewFromObj(vm,
J
Jiri Denemark 已提交
1239 1240
                                     VIR_DOMAIN_EVENT_PMSUSPENDED,
                                     VIR_DOMAIN_EVENT_PMSUSPENDED_MEMORY);
1241

1242
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
1243 1244 1245
            VIR_WARN("Unable to save status on vm %s after suspend event",
                     vm->def->name);
        }
1246 1247 1248

        if (priv->agent)
            qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_SUSPEND);
1249 1250
    }

1251
    virObjectUnlock(vm);
O
Osier Yang 已提交
1252

1253 1254 1255 1256
    if (event)
        qemuDomainEventQueue(driver, event);
    if (lifecycleEvent)
        qemuDomainEventQueue(driver, lifecycleEvent);
1257
    virObjectUnref(cfg);
O
Osier Yang 已提交
1258 1259 1260
    return 0;
}

1261 1262 1263
static int
qemuProcessHandleBalloonChange(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                               virDomainObjPtr vm,
1264 1265
                               unsigned long long actual,
                               void *opaque)
1266
{
1267
    virQEMUDriverPtr driver = opaque;
1268
    virObjectEventPtr event = NULL;
1269
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
1270

1271
    virObjectLock(vm);
1272 1273 1274 1275 1276 1277
    event = virDomainEventBalloonChangeNewFromObj(vm, actual);

    VIR_DEBUG("Updating balloon from %lld to %lld kb",
              vm->def->mem.cur_balloon, actual);
    vm->def->mem.cur_balloon = actual;

1278
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
1279 1280
        VIR_WARN("unable to save domain status with balloon change");

1281
    virObjectUnlock(vm);
1282

1283
    if (event)
1284
        qemuDomainEventQueue(driver, event);
1285
    virObjectUnref(cfg);
1286 1287 1288
    return 0;
}

1289 1290
static int
qemuProcessHandlePMSuspendDisk(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
1291 1292
                               virDomainObjPtr vm,
                               void *opaque)
1293
{
1294
    virQEMUDriverPtr driver = opaque;
1295 1296
    virObjectEventPtr event = NULL;
    virObjectEventPtr lifecycleEvent = NULL;
1297
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
1298

1299
    virObjectLock(vm);
1300 1301 1302 1303 1304 1305 1306 1307 1308 1309
    event = virDomainEventPMSuspendDiskNewFromObj(vm);

    if (virDomainObjGetState(vm, NULL) == VIR_DOMAIN_RUNNING) {
        qemuDomainObjPrivatePtr priv = vm->privateData;
        VIR_DEBUG("Transitioned guest %s to pmsuspended state due to "
                  "QMP suspend_disk event", vm->def->name);

        virDomainObjSetState(vm, VIR_DOMAIN_PMSUSPENDED,
                             VIR_DOMAIN_PMSUSPENDED_UNKNOWN);
        lifecycleEvent =
1310
            virDomainEventLifecycleNewFromObj(vm,
1311 1312 1313
                                     VIR_DOMAIN_EVENT_PMSUSPENDED,
                                     VIR_DOMAIN_EVENT_PMSUSPENDED_DISK);

1314
        if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0) {
1315 1316 1317 1318 1319 1320 1321 1322
            VIR_WARN("Unable to save status on vm %s after suspend event",
                     vm->def->name);
        }

        if (priv->agent)
            qemuAgentNotifyEvent(priv->agent, QEMU_AGENT_EVENT_SUSPEND);
    }

1323
    virObjectUnlock(vm);
1324

1325 1326 1327 1328
    if (event)
        qemuDomainEventQueue(driver, event);
    if (lifecycleEvent)
        qemuDomainEventQueue(driver, lifecycleEvent);
1329 1330
    virObjectUnref(cfg);

1331 1332 1333
    return 0;
}

1334

1335 1336
static int
qemuProcessHandleGuestPanic(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
1337 1338
                            virDomainObjPtr vm,
                            void *opaque)
1339
{
1340
    virQEMUDriverPtr driver = opaque;
1341 1342 1343
    struct qemuProcessEvent *processEvent;

    virObjectLock(vm);
1344
    if (VIR_ALLOC(processEvent) < 0)
1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359
        goto cleanup;

    processEvent->eventType = QEMU_PROCESS_EVENT_GUESTPANIC;
    processEvent->action = vm->def->onCrash;
    processEvent->vm = vm;
    /* Hold an extra reference because we can't allow 'vm' to be
     * deleted before handling guest panic event is finished.
     */
    virObjectRef(vm);
    if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
        if (!virObjectUnref(vm))
            vm = NULL;
        VIR_FREE(processEvent);
    }

1360
 cleanup:
1361
    if (vm)
1362
        virObjectUnlock(vm);
1363 1364 1365 1366 1367

    return 0;
}


1368
int
1369 1370
qemuProcessHandleDeviceDeleted(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                               virDomainObjPtr vm,
1371 1372
                               const char *devAlias,
                               void *opaque)
1373
{
1374
    virQEMUDriverPtr driver = opaque;
1375 1376
    struct qemuProcessEvent *processEvent = NULL;
    char *data;
1377 1378 1379 1380 1381 1382

    virObjectLock(vm);

    VIR_DEBUG("Device %s removed from domain %p %s",
              devAlias, vm, vm->def->name);

1383 1384
    if (qemuDomainSignalDeviceRemoval(vm, devAlias))
        goto cleanup;
1385

1386 1387
    if (VIR_ALLOC(processEvent) < 0)
        goto error;
1388

1389 1390 1391 1392 1393
    processEvent->eventType = QEMU_PROCESS_EVENT_DEVICE_DELETED;
    if (VIR_STRDUP(data, devAlias) < 0)
        goto error;
    processEvent->data = data;
    processEvent->vm = vm;
1394

1395 1396 1397 1398 1399
    virObjectRef(vm);
    if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
        ignore_value(virObjectUnref(vm));
        goto error;
    }
1400

1401
 cleanup:
1402 1403
    virObjectUnlock(vm);
    return 0;
1404 1405 1406 1407 1408
 error:
    if (processEvent)
        VIR_FREE(processEvent->data);
    VIR_FREE(processEvent);
    goto cleanup;
1409 1410 1411
}


1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452
static int
qemuProcessHandleNicRxFilterChanged(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                                    virDomainObjPtr vm,
                                    const char *devAlias,
                                    void *opaque)
{
    virQEMUDriverPtr driver = opaque;
    struct qemuProcessEvent *processEvent = NULL;
    char *data;

    virObjectLock(vm);

    VIR_DEBUG("Device %s RX Filter changed in domain %p %s",
              devAlias, vm, vm->def->name);

    if (VIR_ALLOC(processEvent) < 0)
        goto error;

    processEvent->eventType = QEMU_PROCESS_EVENT_NIC_RX_FILTER_CHANGED;
    if (VIR_STRDUP(data, devAlias) < 0)
        goto error;
    processEvent->data = data;
    processEvent->vm = vm;

    virObjectRef(vm);
    if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
        ignore_value(virObjectUnref(vm));
        goto error;
    }

 cleanup:
    virObjectUnlock(vm);
    return 0;
 error:
    if (processEvent)
        VIR_FREE(processEvent->data);
    VIR_FREE(processEvent);
    goto cleanup;
}


1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495
static int
qemuProcessHandleSerialChanged(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                               virDomainObjPtr vm,
                               const char *devAlias,
                               bool connected,
                               void *opaque)
{
    virQEMUDriverPtr driver = opaque;
    struct qemuProcessEvent *processEvent = NULL;
    char *data;

    virObjectLock(vm);

    VIR_DEBUG("Serial port %s state changed to '%d' in domain %p %s",
              devAlias, connected, vm, vm->def->name);

    if (VIR_ALLOC(processEvent) < 0)
        goto error;

    processEvent->eventType = QEMU_PROCESS_EVENT_SERIAL_CHANGED;
    if (VIR_STRDUP(data, devAlias) < 0)
        goto error;
    processEvent->data = data;
    processEvent->action = connected;
    processEvent->vm = vm;

    virObjectRef(vm);
    if (virThreadPoolSendJob(driver->workerPool, 0, processEvent) < 0) {
        ignore_value(virObjectUnref(vm));
        goto error;
    }

 cleanup:
    virObjectUnlock(vm);
    return 0;
 error:
    if (processEvent)
        VIR_FREE(processEvent->data);
    VIR_FREE(processEvent);
    goto cleanup;
}


1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514
static int
qemuProcessHandleSpiceMigrated(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                               virDomainObjPtr vm,
                               void *opaque ATTRIBUTE_UNUSED)
{
    qemuDomainObjPrivatePtr priv;

    virObjectLock(vm);

    VIR_DEBUG("Spice migration completed for domain %p %s",
              vm, vm->def->name);

    priv = vm->privateData;
    if (priv->job.asyncJob != QEMU_ASYNC_JOB_MIGRATION_OUT) {
        VIR_DEBUG("got SPICE_MIGRATE_COMPLETED event without a migration job");
        goto cleanup;
    }

    priv->job.spiceMigrated = true;
1515
    virDomainObjBroadcast(vm);
1516 1517 1518 1519 1520 1521 1522

 cleanup:
    virObjectUnlock(vm);
    return 0;
}


1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537
static int
qemuProcessHandleMigrationStatus(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                                 virDomainObjPtr vm,
                                 int status,
                                 void *opaque ATTRIBUTE_UNUSED)
{
    qemuDomainObjPrivatePtr priv;

    virObjectLock(vm);

    VIR_DEBUG("Migration of domain %p %s changed state to %s",
              vm, vm->def->name,
              qemuMonitorMigrationStatusTypeToString(status));

    priv = vm->privateData;
1538
    if (priv->job.asyncJob == QEMU_ASYNC_JOB_NONE) {
1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551
        VIR_DEBUG("got MIGRATION event without a migration job");
        goto cleanup;
    }

    priv->job.current->status.status = status;
    virDomainObjBroadcast(vm);

 cleanup:
    virObjectUnlock(vm);
    return 0;
}


1552 1553
static qemuMonitorCallbacks monitorCallbacks = {
    .eofNotify = qemuProcessHandleMonitorEOF,
1554
    .errorNotify = qemuProcessHandleMonitorError,
1555
    .diskSecretLookup = qemuProcessFindVolumeQcowPassphrase,
1556
    .domainEvent = qemuProcessHandleEvent,
1557 1558
    .domainShutdown = qemuProcessHandleShutdown,
    .domainStop = qemuProcessHandleStop,
1559
    .domainResume = qemuProcessHandleResume,
1560 1561 1562 1563 1564
    .domainReset = qemuProcessHandleReset,
    .domainRTCChange = qemuProcessHandleRTCChange,
    .domainWatchdog = qemuProcessHandleWatchdog,
    .domainIOError = qemuProcessHandleIOError,
    .domainGraphics = qemuProcessHandleGraphics,
1565
    .domainBlockJob = qemuProcessHandleBlockJob,
1566
    .domainTrayChange = qemuProcessHandleTrayChange,
O
Osier Yang 已提交
1567
    .domainPMWakeup = qemuProcessHandlePMWakeup,
O
Osier Yang 已提交
1568
    .domainPMSuspend = qemuProcessHandlePMSuspend,
1569
    .domainBalloonChange = qemuProcessHandleBalloonChange,
1570
    .domainPMSuspendDisk = qemuProcessHandlePMSuspendDisk,
1571
    .domainGuestPanic = qemuProcessHandleGuestPanic,
1572
    .domainDeviceDeleted = qemuProcessHandleDeviceDeleted,
1573
    .domainNicRxFilterChanged = qemuProcessHandleNicRxFilterChanged,
1574
    .domainSerialChange = qemuProcessHandleSerialChanged,
1575
    .domainSpiceMigrated = qemuProcessHandleSpiceMigrated,
1576
    .domainMigrationStatus = qemuProcessHandleMigrationStatus,
1577 1578 1579
};

static int
1580 1581
qemuConnectMonitor(virQEMUDriverPtr driver, virDomainObjPtr vm, int asyncJob,
                   int logfd)
1582 1583 1584
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int ret = -1;
1585
    qemuMonitorPtr mon = NULL;
1586

1587
    if (virSecurityManagerSetDaemonSocketLabel(driver->securityManager,
1588
                                               vm->def) < 0) {
1589 1590
        VIR_ERROR(_("Failed to set security context for monitor for %s"),
                  vm->def->name);
1591
        return -1;
1592 1593 1594
    }

    /* Hold an extra reference because we can't allow 'vm' to be
M
Michal Privoznik 已提交
1595
     * deleted unitl the monitor gets its own reference. */
1596
    virObjectRef(vm);
1597

1598
    ignore_value(virTimeMillisNow(&priv->monStart));
1599
    virObjectUnlock(vm);
1600 1601 1602 1603

    mon = qemuMonitorOpen(vm,
                          priv->monConfig,
                          priv->monJSON,
1604 1605
                          &monitorCallbacks,
                          driver);
1606

1607 1608 1609
    if (mon)
        ignore_value(qemuMonitorSetDomainLog(mon, logfd));

1610
    virObjectLock(vm);
M
Michal Privoznik 已提交
1611
    virObjectUnref(vm);
1612
    priv->monStart = 0;
1613

M
Michal Privoznik 已提交
1614
    if (!virDomainObjIsActive(vm)) {
1615
        qemuMonitorClose(mon);
1616
        mon = NULL;
1617 1618 1619
    }
    priv->mon = mon;

1620
    if (virSecurityManagerClearSocketLabel(driver->securityManager, vm->def) < 0) {
1621 1622
        VIR_ERROR(_("Failed to clear security context for monitor for %s"),
                  vm->def->name);
1623
        return -1;
1624 1625 1626 1627
    }

    if (priv->mon == NULL) {
        VIR_INFO("Failed to connect monitor for %s", vm->def->name);
1628
        return -1;
1629 1630 1631
    }


1632
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
1633
        return -1;
1634

1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648
    if (qemuMonitorSetCapabilities(priv->mon) < 0)
        goto cleanup;

    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MONITOR_JSON) &&
        virQEMUCapsProbeQMP(priv->qemuCaps, priv->mon) < 0)
        goto cleanup;

    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT) &&
        qemuMonitorSetMigrationCapability(priv->mon,
                                          QEMU_MONITOR_MIGRATION_CAPS_EVENTS,
                                          true) < 0) {
        VIR_DEBUG("Cannot enable migration events; clearing capability");
        virQEMUCapsClear(priv->qemuCaps, QEMU_CAPS_MIGRATION_EVENT);
    }
1649

1650 1651 1652 1653 1654
    ret = 0;

 cleanup:
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
1655 1656 1657
    return ret;
}

1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671

/**
 * qemuProcessReadLog: Read log file of a qemu VM
 * @fd: File descriptor of the log file
 * @buf: buffer to store the read messages
 * @buflen: allocated space available in @buf
 * @off: Offset to start reading from
 * @skipchar: Skip messages about created character devices
 *
 * Reads log of a qemu VM. Skips messages not produced by qemu or irrelevant
 * messages. Returns length of the message stored in @buf, or -1 on error.
 */
int
qemuProcessReadLog(int fd, char *buf, int buflen, int off, bool skipchar)
1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684
{
    char *filter_next = buf;
    ssize_t bytes;
    char *eol;

    while (off < buflen - 1) {
        bytes = saferead(fd, buf + off, buflen - off - 1);
        if (bytes < 0)
            return -1;

        off += bytes;
        buf[off] = '\0';

1685 1686 1687
        if (bytes == 0)
            break;

1688 1689 1690
        /* Filter out debug messages from intermediate libvirt process */
        while ((eol = strchr(filter_next, '\n'))) {
            *eol = '\0';
1691 1692 1693
            if (virLogProbablyLogMessage(filter_next) ||
                (skipchar &&
                 STRPREFIX(filter_next, "char device redirected to"))) {
1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705
                memmove(filter_next, eol + 1, off - (eol - buf));
                off -= eol + 1 - filter_next;
            } else {
                filter_next = eol + 1;
                *eol = '\n';
            }
        }
    }

    return off;
}

1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723
typedef int qemuProcessLogHandleOutput(virDomainObjPtr vm,
                                       const char *output,
                                       int fd);

/*
 * Returns -1 for error, 0 on success
 */
static int
qemuProcessReadLogOutput(virDomainObjPtr vm,
                         int fd,
                         char *buf,
                         size_t buflen,
                         qemuProcessLogHandleOutput func,
                         const char *what,
                         int timeout)
{
    int retries = (timeout*10);
    int got = 0;
1724 1725
    int ret = -1;

1726 1727 1728
    buf[0] = '\0';

    while (retries) {
1729
        ssize_t func_ret;
1730
        bool isdead;
1731 1732 1733

        func_ret = func(vm, buf, fd);

1734
        isdead = kill(vm->pid, 0) == -1 && errno == ESRCH;
1735

1736
        got = qemuProcessReadLog(fd, buf, buflen, got, false);
1737
        if (got < 0) {
1738 1739 1740
            virReportSystemError(errno,
                                 _("Failure while reading %s log output"),
                                 what);
1741
            goto cleanup;
1742 1743 1744
        }

        if (got == buflen-1) {
1745 1746 1747
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Out of space while reading %s log output: %s"),
                           what, buf);
1748
            goto cleanup;
1749 1750 1751
        }

        if (isdead) {
1752 1753 1754
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Process exited while reading %s log output: %s"),
                           what, buf);
1755
            goto cleanup;
1756 1757
        }

1758 1759 1760 1761
        if (func_ret <= 0) {
            ret = func_ret;
            goto cleanup;
        }
1762 1763 1764 1765 1766

        usleep(100*1000);
        retries--;
    }

1767 1768 1769
    virReportError(VIR_ERR_INTERNAL_ERROR,
                   _("Timed out while reading %s log output: %s"),
                   what, buf);
1770

1771
 cleanup:
1772
    return ret;
1773 1774 1775
}


1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833
/*
 * Read domain log and probably overwrite error if there's one in
 * the domain log file. This function exists to cover the small
 * window between fork() and exec() during which child may fail
 * by libvirt's hand, e.g. placing onto a NUMA node failed.
 */
static int
qemuProcessReadChildErrors(virQEMUDriverPtr driver,
                           virDomainObjPtr vm,
                           off_t originalOff)
{
    int ret = -1;
    int logfd;
    off_t off = 0;
    ssize_t bytes;
    char buf[1024] = {0};
    char *eol, *filter_next = buf;

    if ((logfd = qemuDomainOpenLog(driver, vm, originalOff)) < 0)
        goto cleanup;

    while (off < sizeof(buf) - 1) {
        bytes = saferead(logfd, buf + off, sizeof(buf) - off - 1);
        if (bytes < 0) {
            VIR_WARN("unable to read from log file: %s",
                     virStrerror(errno, buf, sizeof(buf)));
            goto cleanup;
        }

        off += bytes;
        buf[off] = '\0';

        if (bytes == 0)
            break;

        while ((eol = strchr(filter_next, '\n'))) {
            *eol = '\0';
            if (STRPREFIX(filter_next, "libvirt: ")) {
                filter_next = eol + 1;
                *eol = '\n';
                break;
            } else {
                memmove(filter_next, eol + 1, off - (eol - buf));
                off -= eol + 1 - filter_next;
            }
        }
    }

    if (off > 0) {
        /* Found an error in the log. Report it */
        virResetLastError();
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Process exited prior to exec: %s"),
                       buf);
    }

    ret = 0;

1834
 cleanup:
1835 1836 1837 1838 1839
    VIR_FORCE_CLOSE(logfd);
    return ret;
}


1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852
/*
 * Look at a chunk of data from the QEMU stdout logs and try to
 * find a TTY device, as indicated by a line like
 *
 * char device redirected to /dev/pts/3
 *
 * Returns -1 for error, 0 success, 1 continue reading
 */
static int
qemuProcessExtractTTYPath(const char *haystack,
                          size_t *offset,
                          char **path)
{
1853 1854
    static const char needle[] = "char device redirected to";
    char *tmp, *dev;
1855 1856 1857

    VIR_FREE(*path);
    /* First look for our magic string */
1858
    if (!(tmp = strstr(haystack + *offset, needle)))
1859
        return 1;
1860
    tmp += sizeof(needle);
1861 1862 1863 1864 1865 1866 1867 1868
    dev = tmp;

    /*
     * And look for first whitespace character and nul terminate
     * to mark end of the pty path
     */
    while (*tmp) {
        if (c_isspace(*tmp)) {
1869
            if (VIR_STRNDUP(*path, dev, tmp - dev) < 0)
1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887
                return -1;

            /* ... now further update offset till we get EOL */
            *offset = tmp - haystack;
            return 0;
        }
        tmp++;
    }

    /*
     * We found a path, but didn't find any whitespace,
     * so it must be still incomplete - we should at
     * least see a \n - indicate that we want to carry
     * on trying again
     */
    return 1;
}

1888
static int
1889 1890 1891
qemuProcessLookupPTYs(virDomainDefPtr def,
                      virQEMUCapsPtr qemuCaps,
                      virDomainChrDefPtr *devices,
1892
                      int count,
1893
                      virHashTablePtr info)
1894
{
1895
    size_t i;
1896

1897
    for (i = 0; i < count; i++) {
1898
        virDomainChrDefPtr chr = devices[i];
1899 1900
        bool chardevfmt = virQEMUCapsSupportsChardev(def, qemuCaps, chr);

1901
        if (chr->source.type == VIR_DOMAIN_CHR_TYPE_PTY) {
C
Cole Robinson 已提交
1902
            char id[32];
1903
            qemuMonitorChardevInfoPtr entry;
1904

C
Cole Robinson 已提交
1905
            if (snprintf(id, sizeof(id), "%s%s",
1906
                         chardevfmt ? "char" : "",
1907 1908 1909 1910
                         chr->info.alias) >= sizeof(id)) {
                virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                               _("failed to format device alias "
                                 "for PTY retrieval"));
1911
                return -1;
1912
            }
1913

1914 1915
            entry = virHashLookup(info, id);
            if (!entry || !entry->ptyPath) {
1916 1917 1918 1919
                if (chr->source.data.file.path == NULL) {
                    /* neither the log output nor 'info chardev' had a
                     * pty path for this chardev, report an error
                     */
1920 1921
                    virReportError(VIR_ERR_INTERNAL_ERROR,
                                   _("no assigned pty for device %s"), id);
1922 1923 1924 1925 1926 1927 1928 1929 1930 1931
                    return -1;
                } else {
                    /* 'info chardev' had no pty path for this chardev,
                     * but the log output had, so we're fine
                     */
                    continue;
                }
            }

            VIR_FREE(chr->source.data.file.path);
1932
            if (VIR_STRDUP(chr->source.data.file.path, entry->ptyPath) < 0)
1933 1934 1935 1936 1937 1938 1939
                return -1;
        }
    }

    return 0;
}

1940 1941
static int
qemuProcessFindCharDevicePTYsMonitor(virDomainObjPtr vm,
1942
                                     virQEMUCapsPtr qemuCaps,
1943
                                     virHashTablePtr info)
1944
{
1945
    size_t i = 0;
C
Cole Robinson 已提交
1946

1947 1948
    if (qemuProcessLookupPTYs(vm->def, qemuCaps,
                              vm->def->serials, vm->def->nserials,
1949
                              info) < 0)
1950 1951
        return -1;

1952 1953
    if (qemuProcessLookupPTYs(vm->def, qemuCaps,
                              vm->def->parallels, vm->def->nparallels,
1954
                              info) < 0)
1955
        return -1;
1956

1957 1958
    if (qemuProcessLookupPTYs(vm->def, qemuCaps,
                              vm->def->channels, vm->def->nchannels,
1959
                              info) < 0)
1960
        return -1;
1961 1962 1963 1964
    /* For historical reasons, console[0] can be just an alias
     * for serial[0]. That's why we need to update it as well. */
    if (vm->def->nconsoles) {
        virDomainChrDefPtr chr = vm->def->consoles[0];
1965

1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976
        if (vm->def->nserials &&
            chr->deviceType == VIR_DOMAIN_CHR_DEVICE_TYPE_CONSOLE &&
            chr->targetType == VIR_DOMAIN_CHR_CONSOLE_TARGET_TYPE_SERIAL) {
            /* yes, the first console is just an alias for serials[0] */
            i = 1;
            if (virDomainChrSourceDefCopy(&chr->source,
                                          &((vm->def->serials[0])->source)) < 0)
                return -1;
        }
    }

1977 1978
    if (qemuProcessLookupPTYs(vm->def, qemuCaps,
                              vm->def->consoles + i, vm->def->nconsoles - i,
1979
                              info) < 0)
1980
        return -1;
1981 1982 1983 1984 1985 1986 1987 1988 1989 1990

    return 0;
}

static int
qemuProcessFindCharDevicePTYs(virDomainObjPtr vm,
                              const char *output,
                              int fd ATTRIBUTE_UNUSED)
{
    size_t offset = 0;
1991 1992
    int ret;
    size_t i;
1993 1994 1995 1996 1997 1998

    /* The order in which QEMU prints out the PTY paths is
       the order in which it procsses its serial and parallel
       device args. This code must match that ordering.... */

    /* first comes the serial devices */
1999
    for (i = 0; i < vm->def->nserials; i++) {
2000 2001 2002 2003 2004 2005 2006 2007 2008
        virDomainChrDefPtr chr = vm->def->serials[i];
        if (chr->source.type == VIR_DOMAIN_CHR_TYPE_PTY) {
            if ((ret = qemuProcessExtractTTYPath(output, &offset,
                                                 &chr->source.data.file.path)) != 0)
                return ret;
        }
    }

    /* then the parallel devices */
2009
    for (i = 0; i < vm->def->nparallels; i++) {
2010 2011 2012 2013 2014 2015 2016 2017 2018
        virDomainChrDefPtr chr = vm->def->parallels[i];
        if (chr->source.type == VIR_DOMAIN_CHR_TYPE_PTY) {
            if ((ret = qemuProcessExtractTTYPath(output, &offset,
                                                 &chr->source.data.file.path)) != 0)
                return ret;
        }
    }

    /* then the channel devices */
2019
    for (i = 0; i < vm->def->nchannels; i++) {
2020 2021 2022 2023 2024 2025 2026 2027
        virDomainChrDefPtr chr = vm->def->channels[i];
        if (chr->source.type == VIR_DOMAIN_CHR_TYPE_PTY) {
            if ((ret = qemuProcessExtractTTYPath(output, &offset,
                                                 &chr->source.data.file.path)) != 0)
                return ret;
        }
    }

2028
    for (i = 0; i < vm->def->nconsoles; i++) {
2029
        virDomainChrDefPtr chr = vm->def->consoles[i];
2030 2031 2032 2033 2034 2035 2036
        /* For historical reasons, console[0] can be just an alias
         * for serial[0]; That's why we need to update it as well */
        if (i == 0 && vm->def->nserials &&
            chr->deviceType == VIR_DOMAIN_CHR_DEVICE_TYPE_CONSOLE &&
            chr->targetType == VIR_DOMAIN_CHR_CONSOLE_TARGET_TYPE_SERIAL) {
            if ((ret = virDomainChrSourceDefCopy(&chr->source,
                                                 &((vm->def->serials[0])->source))) != 0)
2037
                return ret;
2038 2039 2040 2041 2042 2043 2044
        } else {
            if (chr->source.type == VIR_DOMAIN_CHR_TYPE_PTY &&
                chr->targetType == VIR_DOMAIN_CHR_CONSOLE_TARGET_TYPE_VIRTIO) {
                if ((ret = qemuProcessExtractTTYPath(output, &offset,
                                                     &chr->source.data.file.path)) != 0)
                    return ret;
            }
2045 2046 2047
        }
    }

2048 2049 2050 2051
    return 0;
}


2052
static int
2053 2054 2055 2056
qemuProcessRefreshChannelVirtioState(virQEMUDriverPtr driver,
                                     virDomainObjPtr vm,
                                     virHashTablePtr info,
                                     int booted)
2057 2058
{
    size_t i;
2059
    int agentReason = VIR_CONNECT_DOMAIN_EVENT_AGENT_LIFECYCLE_REASON_CHANNEL;
2060
    qemuMonitorChardevInfoPtr entry;
2061
    virObjectEventPtr event = NULL;
2062 2063
    char id[32];

2064 2065 2066
    if (booted)
        agentReason = VIR_CONNECT_DOMAIN_EVENT_AGENT_LIFECYCLE_REASON_DOMAIN_STARTED;

2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082
    for (i = 0; i < vm->def->nchannels; i++) {
        virDomainChrDefPtr chr = vm->def->channels[i];
        if (chr->targetType == VIR_DOMAIN_CHR_CHANNEL_TARGET_TYPE_VIRTIO) {
            if (snprintf(id, sizeof(id), "char%s",
                         chr->info.alias) >= sizeof(id)) {
                virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                               _("failed to format device alias "
                                 "for PTY retrieval"));
                return -1;
            }

            /* port state not reported */
            if (!(entry = virHashLookup(info, id)) ||
                !entry->state)
                continue;

2083 2084 2085 2086 2087 2088
            if (entry->state != VIR_DOMAIN_CHR_DEVICE_STATE_DEFAULT &&
                STREQ_NULLABLE(chr->target.name, "org.qemu.guest_agent.0") &&
                (event = virDomainEventAgentLifecycleNewFromObj(vm, entry->state,
                                                                agentReason)))
                qemuDomainEventQueue(driver, event);

2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106
            chr->state = entry->state;
        }
    }

    return 0;
}


static int
qemuProcessReconnectRefreshChannelVirtioState(virQEMUDriverPtr driver,
                                              virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virHashTablePtr info = NULL;
    int ret = -1;

    qemuDomainObjEnterMonitor(driver, vm);
    ret = qemuMonitorGetChardevInfo(priv->mon, &info);
2107 2108
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
2109 2110 2111 2112

    if (ret < 0)
        goto cleanup;

2113
    ret = qemuProcessRefreshChannelVirtioState(driver, vm, info, false);
2114 2115 2116 2117 2118 2119 2120

 cleanup:
    virHashFree(info);
    return ret;
}


2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152
static int
qemuProcessRefreshBalloonState(virQEMUDriverPtr driver,
                               virDomainObjPtr vm,
                               int asyncJob)
{
    unsigned long long balloon;
    int rc;

    /* if no ballooning is available, the current size equals to the current
     * full memory size */
    if (!vm->def->memballoon ||
        vm->def->memballoon->model == VIR_DOMAIN_MEMBALLOON_MODEL_NONE) {
        vm->def->mem.cur_balloon = virDomainDefGetMemoryActual(vm->def);
        return 0;
    }

    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        return -1;

    rc = qemuMonitorGetBalloonInfo(qemuDomainGetMonitor(vm), &balloon);
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        rc = -1;

    if (rc < 0)
        return -1;

    vm->def->mem.cur_balloon = balloon;

    return 0;
}


2153
static int
2154
qemuProcessWaitForMonitor(virQEMUDriverPtr driver,
C
Cole Robinson 已提交
2155
                          virDomainObjPtr vm,
2156
                          int asyncJob,
2157
                          virQEMUCapsPtr qemuCaps,
C
Cole Robinson 已提交
2158
                          off_t pos)
2159
{
2160
    char *buf = NULL;
2161
    size_t buf_size = 4096; /* Plenty of space to get startup greeting */
2162
    int logfd = -1;
2163
    int ret = -1;
2164
    virHashTablePtr info = NULL;
2165
    qemuDomainObjPrivatePtr priv;
2166

2167 2168 2169
    if (pos != -1 &&
        (logfd = qemuDomainOpenLog(driver, vm, pos)) < 0)
        return -1;
2170

2171
    if (logfd != -1 && !virQEMUCapsUsedQMP(qemuCaps)) {
2172
        if (VIR_ALLOC_N(buf, buf_size) < 0)
2173
            goto closelog;
2174

2175 2176 2177 2178 2179
        if (qemuProcessReadLogOutput(vm, logfd, buf, buf_size,
                                     qemuProcessFindCharDevicePTYs,
                                     "console", 30) < 0)
            goto closelog;
    }
2180 2181

    VIR_DEBUG("Connect monitor to %p '%s'", vm, vm->def->name);
2182
    if (qemuConnectMonitor(driver, vm, asyncJob, logfd) < 0)
2183 2184 2185 2186 2187 2188
        goto cleanup;

    /* Try to get the pty path mappings again via the monitor. This is much more
     * reliable if it's available.
     * Note that the monitor itself can be on a pty, so we still need to try the
     * log output method. */
2189
    priv = vm->privateData;
2190 2191
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto cleanup;
2192 2193
    ret = qemuMonitorGetChardevInfo(priv->mon, &info);
    VIR_DEBUG("qemuMonitorGetChardevInfo returned %i", ret);
2194 2195 2196
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;

2197 2198 2199 2200 2201
    if (ret == 0) {
        if ((ret = qemuProcessFindCharDevicePTYsMonitor(vm, qemuCaps,
                                                        info)) < 0)
            goto cleanup;

2202 2203
        if ((ret = qemuProcessRefreshChannelVirtioState(driver, vm, info,
                                                        true)) < 0)
2204 2205
            goto cleanup;
    }
2206

2207
 cleanup:
2208
    virHashFree(info);
2209

E
Eric Blake 已提交
2210
    if (pos != -1 && kill(vm->pid, 0) == -1 && errno == ESRCH) {
2211
        int len;
2212 2213
        /* VM is dead, any other error raised in the interim is probably
         * not as important as the qemu cmdline output */
2214
        if (virQEMUCapsUsedQMP(qemuCaps)) {
2215
            if (VIR_ALLOC_N(buf, buf_size) < 0)
2216 2217
                goto closelog;
        }
2218 2219

        len = strlen(buf);
2220
        /* best effort seek - we need to reset to the original position, so that
2221 2222
         * a possible read of the fd in the monitor code doesn't influence this
         * error delivery option */
2223
        ignore_value(lseek(logfd, pos, SEEK_SET));
2224
        qemuProcessReadLog(logfd, buf + len, buf_size - len - 1, 0, true);
2225 2226 2227
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("process exited while connecting to monitor: %s"),
                       buf);
2228 2229 2230
        ret = -1;
    }

2231
 closelog:
2232
    if (VIR_CLOSE(logfd) < 0) {
2233
        char ebuf[1024];
2234
        VIR_WARN("Unable to close logfile: %s",
2235
                 virStrerror(errno, ebuf, sizeof(ebuf)));
2236 2237
    }

2238 2239
    VIR_FREE(buf);

2240 2241 2242 2243
    return ret;
}

static int
2244
qemuProcessDetectVcpuPIDs(virQEMUDriverPtr driver,
2245
                          virDomainObjPtr vm, int asyncJob)
2246 2247 2248 2249 2250
{
    pid_t *cpupids = NULL;
    int ncpupids;
    qemuDomainObjPrivatePtr priv = vm->privateData;

2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284
    /*
     * Current QEMU *can* report info about host threads mapped
     * to vCPUs, but it is not in a manner we can correctly
     * deal with. The TCG CPU emulation does have a separate vCPU
     * thread, but it runs every vCPU in that same thread. So it
     * is impossible to setup different affinity per thread.
     *
     * What's more the 'query-cpus' command returns bizarre
     * data for the threads. It gives the TCG thread for the
     * vCPU 0, but for vCPUs 1-> N, it actually replies with
     * the main process thread ID.
     *
     * The result is that when we try to set affinity for
     * vCPU 1, it will actually change the affinity of the
     * emulator thread :-( When you try to set affinity for
     * vCPUs 2, 3.... it will fail if the affinity was
     * different from vCPU 1.
     *
     * We *could* allow vcpu pinning with TCG, if we made the
     * restriction that all vCPUs had the same mask. This would
     * at least let us separate emulator from vCPUs threads, as
     * we do for KVM. It would need some changes to our cgroups
     * CPU layout though, and error reporting for the config
     * restrictions.
     *
     * Just disable CPU pinning with TCG until someone wants
     * to try to do this hard work.
     */
    if (vm->def->virtType == VIR_DOMAIN_VIRT_QEMU) {
        priv->nvcpupids = 0;
        priv->vcpupids = NULL;
        return 0;
    }

2285 2286
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        return -1;
2287 2288 2289
    ncpupids = qemuMonitorGetCPUInfo(priv->mon, &cpupids);
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        return -1;
2290 2291 2292
    /* failure to get the VCPU<-> PID mapping or to execute the query
     * command will not be treated fatal as some versions of qemu don't
     * support this command */
2293
    if (ncpupids <= 0) {
2294 2295
        virResetLastError();

2296 2297
        priv->nvcpupids = 0;
        priv->vcpupids = NULL;
2298 2299 2300 2301
        return 0;
    }

    if (ncpupids != vm->def->vcpus) {
2302 2303 2304 2305
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("got wrong number of vCPU pids from QEMU monitor. "
                         "got %d, wanted %d"),
                       ncpupids, vm->def->vcpus);
2306 2307 2308 2309 2310 2311 2312 2313 2314
        VIR_FREE(cpupids);
        return -1;
    }

    priv->nvcpupids = ncpupids;
    priv->vcpupids = cpupids;
    return 0;
}

2315

2316 2317 2318 2319 2320 2321
static int
qemuProcessDetectIOThreadPIDs(virQEMUDriverPtr driver,
                              virDomainObjPtr vm,
                              int asyncJob)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
2322
    qemuMonitorIOThreadInfoPtr *iothreads = NULL;
2323 2324 2325 2326
    int niothreads = 0;
    int ret = -1;
    size_t i;

2327 2328 2329
    if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_OBJECT_IOTHREAD))
        return 0;

2330 2331 2332 2333
    /* Get the list of IOThreads from qemu */
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto cleanup;
    niothreads = qemuMonitorGetIOThreads(priv->mon, &iothreads);
2334 2335
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        goto cleanup;
J
John Ferlan 已提交
2336
    if (niothreads < 0)
2337 2338 2339 2340 2341 2342 2343 2344 2345 2346
        goto cleanup;

    if (niothreads != vm->def->iothreads) {
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("got wrong number of IOThread pids from QEMU monitor. "
                         "got %d, wanted %d"),
                       niothreads, vm->def->iothreads);
        goto cleanup;
    }

2347 2348 2349 2350 2351 2352
    /* Nothing to do */
    if (niothreads == 0) {
        ret = 0;
        goto cleanup;
    }

2353 2354 2355
    for (i = 0; i < niothreads; i++) {
        virDomainIOThreadIDDefPtr iothrid;

2356 2357
        if (!(iothrid = virDomainIOThreadIDFind(vm->def,
                                                iothreads[i]->iothread_id))) {
2358
            virReportError(VIR_ERR_INTERNAL_ERROR,
2359 2360
                           _("iothread %d not found"),
                           iothreads[i]->iothread_id);
2361 2362 2363 2364
            goto cleanup;
        }
        iothrid->thread_id = iothreads[i]->thread_id;
    }
2365 2366 2367 2368 2369 2370

    ret = 0;

 cleanup:
    if (iothreads) {
        for (i = 0; i < niothreads; i++)
2371
            VIR_FREE(iothreads[i]);
2372 2373 2374 2375 2376
        VIR_FREE(iothreads);
    }
    return ret;
}

2377 2378 2379 2380 2381

/*
 * To be run between fork/exec of QEMU only
 */
static int
2382
qemuProcessInitCpuAffinity(virDomainObjPtr vm)
2383 2384 2385 2386
{
    int ret = -1;
    virBitmapPtr cpumap = NULL;
    virBitmapPtr cpumapToSet = NULL;
2387
    qemuDomainObjPrivatePtr priv = vm->privateData;
2388

2389 2390 2391 2392 2393 2394
    if (!vm->pid) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Cannot setup CPU affinity until process is started"));
        return -1;
    }

2395 2396
    if (vm->def->placement_mode == VIR_DOMAIN_CPU_PLACEMENT_MODE_AUTO) {
        VIR_DEBUG("Set CPU affinity with advisory nodeset from numad");
2397
        cpumapToSet = priv->autoCpuset;
2398
    } else {
2399
        VIR_DEBUG("Set CPU affinity with specified cpuset");
O
Osier Yang 已提交
2400
        if (vm->def->cpumask) {
H
Hu Tao 已提交
2401
            cpumapToSet = vm->def->cpumask;
O
Osier Yang 已提交
2402 2403 2404 2405 2406
        } else {
            /* You may think this is redundant, but we can't assume libvirtd
             * itself is running on all pCPUs, so we need to explicitly set
             * the spawned QEMU instance to all pCPUs if no map is given in
             * its config file */
2407 2408 2409 2410
            int hostcpus;

            /* setaffinity fails if you set bits for CPUs which
             * aren't present, so we have to limit ourselves */
2411
            if ((hostcpus = nodeGetCPUCount(NULL)) < 0)
2412 2413 2414 2415 2416 2417 2418 2419
                goto cleanup;

            if (hostcpus > QEMUD_CPUMASK_LEN)
                hostcpus = QEMUD_CPUMASK_LEN;

            if (!(cpumap = virBitmapNew(hostcpus)))
                goto cleanup;

2420
            virBitmapSetAll(cpumap);
2421 2422

            cpumapToSet = cpumap;
O
Osier Yang 已提交
2423
        }
2424 2425
    }

2426
    if (virProcessSetAffinity(vm->pid, cpumapToSet) < 0)
2427
        goto cleanup;
2428

2429 2430
    ret = 0;

2431
 cleanup:
2432
    virBitmapFree(cpumap);
2433
    return ret;
2434 2435
}

2436 2437 2438 2439 2440 2441
/* set link states to down on interfaces at qemu start */
static int
qemuProcessSetLinkStates(virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainDefPtr def = vm->def;
2442
    size_t i;
2443 2444 2445 2446
    int ret = 0;

    for (i = 0; i < def->nnets; i++) {
        if (def->nets[i]->linkstate == VIR_DOMAIN_NET_INTERFACE_LINK_STATE_DOWN) {
2447 2448 2449 2450 2451 2452
            if (!def->nets[i]->info.alias) {
                virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                               _("missing alias for network device"));
                return -1;
            }

2453 2454
            VIR_DEBUG("Setting link state: %s", def->nets[i]->info.alias);

2455
            if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_NETDEV)) {
2456
                virReportError(VIR_ERR_OPERATION_UNSUPPORTED, "%s",
2457
                               _("Setting of link state is not supported by this qemu"));
2458 2459 2460 2461 2462 2463 2464
                return -1;
            }

            ret = qemuMonitorSetLink(priv->mon,
                                     def->nets[i]->info.alias,
                                     VIR_DOMAIN_NET_INTERFACE_LINK_STATE_DOWN);
            if (ret != 0) {
2465
                virReportError(VIR_ERR_OPERATION_FAILED,
2466 2467 2468 2469 2470 2471 2472 2473 2474 2475
                               _("Couldn't set link state on interface: %s"), def->nets[i]->info.alias);
                break;
            }
        }
    }

    return ret;
}

/* Set CPU affinities for vcpus if vcpupin xml provided. */
2476
static int
2477
qemuProcessSetVcpuAffinities(virDomainObjPtr vm)
2478 2479 2480
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainDefPtr def = vm->def;
2481
    virDomainPinDefPtr pininfo;
2482
    int n;
E
Eric Blake 已提交
2483
    int ret = -1;
2484 2485
    VIR_DEBUG("Setting affinity on CPUs nvcpupin=%zu nvcpus=%d nvcpupids=%d",
              def->cputune.nvcpupin, def->vcpus, priv->nvcpupids);
2486 2487 2488 2489
    if (!def->cputune.nvcpupin)
        return 0;

    if (priv->vcpupids == NULL) {
2490 2491 2492
        /* If any CPU has custom affinity that differs from the
         * VM default affinity, we must reject it
         */
2493
        for (n = 0; n < def->cputune.nvcpupin; n++) {
2494 2495 2496 2497 2498 2499 2500 2501
            if (!virBitmapEqual(def->cpumask,
                                def->cputune.vcpupin[n]->cpumask)) {
                virReportError(VIR_ERR_OPERATION_INVALID,
                               "%s", _("cpu affinity is not supported"));
                return -1;
            }
        }
        return 0;
2502 2503
    }

2504 2505
    for (n = 0; n < def->vcpus; n++) {
        /* set affinity only for existing vcpus */
2506 2507 2508
        if (!(pininfo = virDomainPinFind(def->cputune.vcpupin,
                                         def->cputune.nvcpupin,
                                         n)))
2509
            continue;
2510

2511 2512
        if (virProcessSetAffinity(priv->vcpupids[n],
                                  pininfo->cpumask) < 0) {
E
Eric Blake 已提交
2513
            goto cleanup;
2514 2515 2516
        }
    }

E
Eric Blake 已提交
2517
    ret = 0;
2518
 cleanup:
E
Eric Blake 已提交
2519
    return ret;
2520
}
2521

2522
/* Set CPU affinities for emulator threads. */
2523
static int
2524
qemuProcessSetEmulatorAffinity(virDomainObjPtr vm)
2525
{
2526
    virBitmapPtr cpumask;
2527 2528 2529
    virDomainDefPtr def = vm->def;
    int ret = -1;

O
Osier Yang 已提交
2530
    if (def->cputune.emulatorpin)
2531
        cpumask = def->cputune.emulatorpin;
O
Osier Yang 已提交
2532
    else if (def->cpumask)
2533
        cpumask = def->cpumask;
O
Osier Yang 已提交
2534 2535
    else
        return 0;
2536

2537
    ret = virProcessSetAffinity(vm->pid, cpumask);
2538 2539 2540
    return ret;
}

2541 2542 2543 2544 2545 2546 2547 2548
/* Set CPU affinities for IOThreads threads. */
static int
qemuProcessSetIOThreadsAffinity(virDomainObjPtr vm)
{
    virDomainDefPtr def = vm->def;
    size_t i;
    int ret = -1;

2549 2550
    for (i = 0; i < def->niothreadids; i++) {
        /* set affinity only for existing iothreads */
2551
        if (!def->iothreadids[i]->cpumask)
2552 2553
            continue;

2554
        if (virProcessSetAffinity(def->iothreadids[i]->thread_id,
2555
                                  def->iothreadids[i]->cpumask) < 0)
2556 2557 2558 2559 2560 2561 2562 2563
            goto cleanup;
    }
    ret = 0;

 cleanup:
    return ret;
}

2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588
/* Set Scheduler parameters for vCPU or I/O threads. */
int
qemuProcessSetSchedParams(int id,
                          pid_t pid,
                          size_t nsp,
                          virDomainThreadSchedParamPtr sp)
{
    bool val = false;
    size_t i = 0;
    virDomainThreadSchedParamPtr s = NULL;

    for (i = 0; i < nsp; i++) {
        if (virBitmapGetBit(sp[i].ids, id, &val) < 0) {
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("Cannot get bit from bitmap"));
        }
        if (val) {
            s = &sp[i];
            break;
        }
    }

    if (!s)
        return 0;

2589
    return virProcessSetScheduler(pid, s->policy, s->priority);
2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604
}

static int
qemuProcessSetSchedulers(virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    size_t i = 0;

    for (i = 0; i < priv->nvcpupids; i++) {
        if (qemuProcessSetSchedParams(i, priv->vcpupids[i],
                                      vm->def->cputune.nvcpusched,
                                      vm->def->cputune.vcpusched) < 0)
            return -1;
    }

2605 2606 2607
    for (i = 0; i < vm->def->niothreadids; i++) {
        if (qemuProcessSetSchedParams(vm->def->iothreadids[i]->iothread_id,
                                      vm->def->iothreadids[i]->thread_id,
2608 2609 2610 2611 2612 2613 2614 2615
                                      vm->def->cputune.niothreadsched,
                                      vm->def->cputune.iothreadsched) < 0)
            return -1;
    }

    return 0;
}

2616 2617
static int
qemuProcessInitPasswords(virConnectPtr conn,
2618
                         virQEMUDriverPtr driver,
2619 2620
                         virDomainObjPtr vm,
                         int asyncJob)
2621 2622 2623
{
    int ret = 0;
    qemuDomainObjPrivatePtr priv = vm->privateData;
2624
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
2625
    size_t i;
2626 2627
    char *alias = NULL;
    char *secret = NULL;
2628

2629
    for (i = 0; i < vm->def->ngraphics; ++i) {
2630 2631
        virDomainGraphicsDefPtr graphics = vm->def->graphics[i];
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC) {
2632 2633
            ret = qemuDomainChangeGraphicsPasswords(driver, vm,
                                                    VIR_DOMAIN_GRAPHICS_TYPE_VNC,
2634
                                                    &graphics->data.vnc.auth,
2635 2636
                                                    cfg->vncPassword,
                                                    asyncJob);
2637
        } else if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
2638 2639
            ret = qemuDomainChangeGraphicsPasswords(driver, vm,
                                                    VIR_DOMAIN_GRAPHICS_TYPE_SPICE,
2640
                                                    &graphics->data.spice.auth,
2641 2642
                                                    cfg->spicePassword,
                                                    asyncJob);
2643 2644
        }

2645 2646 2647
        if (ret < 0)
            goto cleanup;
    }
2648

2649
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE)) {
2650
        for (i = 0; i < vm->def->ndisks; i++) {
2651 2652
            size_t secretLen;

2653
            if (!vm->def->disks[i]->src->encryption ||
2654
                !virDomainDiskGetSource(vm->def->disks[i]))
2655 2656
                continue;

2657
            VIR_FREE(secret);
2658 2659 2660 2661 2662
            if (qemuProcessGetVolumeQcowPassphrase(conn,
                                                   vm->def->disks[i],
                                                   &secret, &secretLen) < 0)
                goto cleanup;

2663 2664 2665 2666
            VIR_FREE(alias);
            if (VIR_STRDUP(alias, vm->def->disks[i]->info.alias) < 0)
                goto cleanup;
            if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
2667
                goto cleanup;
E
Eric Blake 已提交
2668
            ret = qemuMonitorSetDrivePassphrase(priv->mon, alias, secret);
2669 2670
            if (qemuDomainObjExitMonitor(driver, vm) < 0)
                ret = -1;
2671 2672 2673 2674 2675
            if (ret < 0)
                goto cleanup;
        }
    }

2676
 cleanup:
2677 2678
    VIR_FREE(alias);
    VIR_FREE(secret);
2679
    virObjectUnref(cfg);
2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722
    return ret;
}


#define QEMU_PCI_VENDOR_INTEL     0x8086
#define QEMU_PCI_VENDOR_LSI_LOGIC 0x1000
#define QEMU_PCI_VENDOR_REDHAT    0x1af4
#define QEMU_PCI_VENDOR_CIRRUS    0x1013
#define QEMU_PCI_VENDOR_REALTEK   0x10ec
#define QEMU_PCI_VENDOR_AMD       0x1022
#define QEMU_PCI_VENDOR_ENSONIQ   0x1274
#define QEMU_PCI_VENDOR_VMWARE    0x15ad
#define QEMU_PCI_VENDOR_QEMU      0x1234

#define QEMU_PCI_PRODUCT_DISK_VIRTIO 0x1001

#define QEMU_PCI_PRODUCT_BALLOON_VIRTIO 0x1002

#define QEMU_PCI_PRODUCT_NIC_NE2K     0x8029
#define QEMU_PCI_PRODUCT_NIC_PCNET    0x2000
#define QEMU_PCI_PRODUCT_NIC_RTL8139  0x8139
#define QEMU_PCI_PRODUCT_NIC_E1000    0x100E
#define QEMU_PCI_PRODUCT_NIC_VIRTIO   0x1000

#define QEMU_PCI_PRODUCT_VGA_CIRRUS 0x00b8
#define QEMU_PCI_PRODUCT_VGA_VMWARE 0x0405
#define QEMU_PCI_PRODUCT_VGA_STDVGA 0x1111

#define QEMU_PCI_PRODUCT_AUDIO_AC97    0x2415
#define QEMU_PCI_PRODUCT_AUDIO_ES1370  0x5000

#define QEMU_PCI_PRODUCT_CONTROLLER_PIIX 0x7010
#define QEMU_PCI_PRODUCT_CONTROLLER_LSI  0x0012

#define QEMU_PCI_PRODUCT_WATCHDOG_I63000ESB 0x25ab

static int
qemuProcessAssignNextPCIAddress(virDomainDeviceInfo *info,
                                int vendor,
                                int product,
                                qemuMonitorPCIAddress *addrs,
                                int naddrs)
{
2723
    bool found = false;
2724
    size_t i;
2725 2726 2727

    VIR_DEBUG("Look for %x:%x out of %d", vendor, product, naddrs);

2728
    for (i = 0; i < naddrs; i++) {
2729 2730 2731
        VIR_DEBUG("Maybe %x:%x", addrs[i].vendor, addrs[i].product);
        if (addrs[i].vendor == vendor &&
            addrs[i].product == product) {
2732
            VIR_DEBUG("Match %zu", i);
2733
            found = true;
2734 2735 2736
            break;
        }
    }
2737
    if (!found)
2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937
        return -1;

    /* Blank it out so this device isn't matched again */
    addrs[i].vendor = 0;
    addrs[i].product = 0;

    if (info->type == VIR_DOMAIN_DEVICE_ADDRESS_TYPE_NONE)
        info->type = VIR_DOMAIN_DEVICE_ADDRESS_TYPE_PCI;

    if (info->type == VIR_DOMAIN_DEVICE_ADDRESS_TYPE_PCI) {
        info->addr.pci.domain = addrs[i].addr.domain;
        info->addr.pci.bus = addrs[i].addr.bus;
        info->addr.pci.slot = addrs[i].addr.slot;
        info->addr.pci.function = addrs[i].addr.function;
    }

    return 0;
}

static int
qemuProcessGetPCIDiskVendorProduct(virDomainDiskDefPtr def,
                                   unsigned *vendor,
                                   unsigned *product)
{
    switch (def->bus) {
    case VIR_DOMAIN_DISK_BUS_VIRTIO:
        *vendor = QEMU_PCI_VENDOR_REDHAT;
        *product = QEMU_PCI_PRODUCT_DISK_VIRTIO;
        break;

    default:
        return -1;
    }

    return 0;
}

static int
qemuProcessGetPCINetVendorProduct(virDomainNetDefPtr def,
                                  unsigned *vendor,
                                  unsigned *product)
{
    if (!def->model)
        return -1;

    if (STREQ(def->model, "ne2k_pci")) {
        *vendor = QEMU_PCI_VENDOR_REALTEK;
        *product = QEMU_PCI_PRODUCT_NIC_NE2K;
    } else if (STREQ(def->model, "pcnet")) {
        *vendor = QEMU_PCI_VENDOR_AMD;
        *product = QEMU_PCI_PRODUCT_NIC_PCNET;
    } else if (STREQ(def->model, "rtl8139")) {
        *vendor = QEMU_PCI_VENDOR_REALTEK;
        *product = QEMU_PCI_PRODUCT_NIC_RTL8139;
    } else if (STREQ(def->model, "e1000")) {
        *vendor = QEMU_PCI_VENDOR_INTEL;
        *product = QEMU_PCI_PRODUCT_NIC_E1000;
    } else if (STREQ(def->model, "virtio")) {
        *vendor = QEMU_PCI_VENDOR_REDHAT;
        *product = QEMU_PCI_PRODUCT_NIC_VIRTIO;
    } else {
        VIR_INFO("Unexpected NIC model %s, cannot get PCI address",
                 def->model);
        return -1;
    }
    return 0;
}

static int
qemuProcessGetPCIControllerVendorProduct(virDomainControllerDefPtr def,
                                         unsigned *vendor,
                                         unsigned *product)
{
    switch (def->type) {
    case VIR_DOMAIN_CONTROLLER_TYPE_SCSI:
        *vendor = QEMU_PCI_VENDOR_LSI_LOGIC;
        *product = QEMU_PCI_PRODUCT_CONTROLLER_LSI;
        break;

    case VIR_DOMAIN_CONTROLLER_TYPE_FDC:
        /* XXX we could put in the ISA bridge address, but
           that's not technically the FDC's address */
        return -1;

    case VIR_DOMAIN_CONTROLLER_TYPE_IDE:
        *vendor = QEMU_PCI_VENDOR_INTEL;
        *product = QEMU_PCI_PRODUCT_CONTROLLER_PIIX;
        break;

    default:
        VIR_INFO("Unexpected controller type %s, cannot get PCI address",
                 virDomainControllerTypeToString(def->type));
        return -1;
    }

    return 0;
}

static int
qemuProcessGetPCIVideoVendorProduct(virDomainVideoDefPtr def,
                                    unsigned *vendor,
                                    unsigned *product)
{
    switch (def->type) {
    case VIR_DOMAIN_VIDEO_TYPE_CIRRUS:
        *vendor = QEMU_PCI_VENDOR_CIRRUS;
        *product = QEMU_PCI_PRODUCT_VGA_CIRRUS;
        break;

    case VIR_DOMAIN_VIDEO_TYPE_VGA:
        *vendor = QEMU_PCI_VENDOR_QEMU;
        *product = QEMU_PCI_PRODUCT_VGA_STDVGA;
        break;

    case VIR_DOMAIN_VIDEO_TYPE_VMVGA:
        *vendor = QEMU_PCI_VENDOR_VMWARE;
        *product = QEMU_PCI_PRODUCT_VGA_VMWARE;
        break;

    default:
        return -1;
    }
    return 0;
}

static int
qemuProcessGetPCISoundVendorProduct(virDomainSoundDefPtr def,
                                    unsigned *vendor,
                                    unsigned *product)
{
    switch (def->model) {
    case VIR_DOMAIN_SOUND_MODEL_ES1370:
        *vendor = QEMU_PCI_VENDOR_ENSONIQ;
        *product = QEMU_PCI_PRODUCT_AUDIO_ES1370;
        break;

    case VIR_DOMAIN_SOUND_MODEL_AC97:
        *vendor = QEMU_PCI_VENDOR_INTEL;
        *product = QEMU_PCI_PRODUCT_AUDIO_AC97;
        break;

    default:
        return -1;
    }

    return 0;
}

static int
qemuProcessGetPCIWatchdogVendorProduct(virDomainWatchdogDefPtr def,
                                       unsigned *vendor,
                                       unsigned *product)
{
    switch (def->model) {
    case VIR_DOMAIN_WATCHDOG_MODEL_I6300ESB:
        *vendor = QEMU_PCI_VENDOR_INTEL;
        *product = QEMU_PCI_PRODUCT_WATCHDOG_I63000ESB;
        break;

    default:
        return -1;
    }

    return 0;
}


static int
qemuProcessGetPCIMemballoonVendorProduct(virDomainMemballoonDefPtr def,
                                         unsigned *vendor,
                                         unsigned *product)
{
    switch (def->model) {
    case VIR_DOMAIN_MEMBALLOON_MODEL_VIRTIO:
        *vendor = QEMU_PCI_VENDOR_REDHAT;
        *product = QEMU_PCI_PRODUCT_BALLOON_VIRTIO;
        break;

    default:
        return -1;
    }

    return 0;
}


/*
 * This entire method assumes that PCI devices in 'info pci'
 * match ordering of devices specified on the command line
 * wrt to devices of matching vendor+product
 *
 * XXXX this might not be a valid assumption if we assign
 * some static addrs on CLI. Have to check that...
 */
static int
qemuProcessDetectPCIAddresses(virDomainObjPtr vm,
                              qemuMonitorPCIAddress *addrs,
                              int naddrs)
{
    unsigned int vendor = 0, product = 0;
2938
    size_t i;
2939 2940 2941 2942 2943

    /* XXX should all these vendor/product IDs be kept in the
     * actual device data structure instead ?
     */

2944
    for (i = 0; i < vm->def->ndisks; i++) {
2945 2946 2947 2948 2949 2950
        if (qemuProcessGetPCIDiskVendorProduct(vm->def->disks[i], &vendor, &product) < 0)
            continue;

        if (qemuProcessAssignNextPCIAddress(&(vm->def->disks[i]->info),
                                            vendor, product,
                                            addrs, naddrs) < 0) {
2951 2952 2953
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for VirtIO disk %s"),
                           vm->def->disks[i]->dst);
2954 2955 2956 2957
            return -1;
        }
    }

2958
    for (i = 0; i < vm->def->nnets; i++) {
2959 2960 2961 2962 2963 2964
        if (qemuProcessGetPCINetVendorProduct(vm->def->nets[i], &vendor, &product) < 0)
            continue;

        if (qemuProcessAssignNextPCIAddress(&(vm->def->nets[i]->info),
                                            vendor, product,
                                            addrs,  naddrs) < 0) {
2965 2966 2967
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for %s NIC"),
                           vm->def->nets[i]->model);
2968 2969 2970 2971
            return -1;
        }
    }

2972
    for (i = 0; i < vm->def->ncontrollers; i++) {
2973 2974 2975 2976 2977 2978
        if (qemuProcessGetPCIControllerVendorProduct(vm->def->controllers[i], &vendor, &product) < 0)
            continue;

        if (qemuProcessAssignNextPCIAddress(&(vm->def->controllers[i]->info),
                                            vendor, product,
                                            addrs,  naddrs) < 0) {
2979 2980 2981
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for controller %s"),
                           virDomainControllerTypeToString(vm->def->controllers[i]->type));
2982 2983 2984 2985
            return -1;
        }
    }

2986
    for (i = 0; i < vm->def->nvideos; i++) {
2987 2988 2989 2990 2991 2992
        if (qemuProcessGetPCIVideoVendorProduct(vm->def->videos[i], &vendor, &product) < 0)
            continue;

        if (qemuProcessAssignNextPCIAddress(&(vm->def->videos[i]->info),
                                            vendor, product,
                                            addrs,  naddrs) < 0) {
2993 2994 2995
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for video adapter %s"),
                           virDomainVideoTypeToString(vm->def->videos[i]->type));
2996 2997 2998 2999
            return -1;
        }
    }

3000
    for (i = 0; i < vm->def->nsounds; i++) {
3001 3002 3003 3004 3005 3006
        if (qemuProcessGetPCISoundVendorProduct(vm->def->sounds[i], &vendor, &product) < 0)
            continue;

        if (qemuProcessAssignNextPCIAddress(&(vm->def->sounds[i]->info),
                                    vendor, product,
                                     addrs,  naddrs) < 0) {
3007 3008 3009
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for sound adapter %s"),
                           virDomainSoundModelTypeToString(vm->def->sounds[i]->model));
3010 3011 3012 3013 3014 3015 3016 3017 3018 3019
            return -1;
        }
    }


    if (vm->def->watchdog &&
        qemuProcessGetPCIWatchdogVendorProduct(vm->def->watchdog, &vendor, &product) == 0) {
        if (qemuProcessAssignNextPCIAddress(&(vm->def->watchdog->info),
                                            vendor, product,
                                            addrs,  naddrs) < 0) {
3020 3021 3022
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for watchdog %s"),
                           virDomainWatchdogModelTypeToString(vm->def->watchdog->model));
3023 3024 3025 3026 3027 3028 3029 3030 3031
            return -1;
        }
    }

    if (vm->def->memballoon &&
        qemuProcessGetPCIMemballoonVendorProduct(vm->def->memballoon, &vendor, &product) == 0) {
        if (qemuProcessAssignNextPCIAddress(&(vm->def->memballoon->info),
                                            vendor, product,
                                            addrs, naddrs) < 0) {
3032 3033 3034
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("cannot find PCI address for balloon %s"),
                           virDomainMemballoonModelTypeToString(vm->def->memballoon->model));
3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051
            return -1;
        }
    }

    /* XXX console (virtio) */


    /* ... and now things we don't have in our xml */

    /* XXX USB controller ? */

    /* XXX what about other PCI devices (ie bridges) */

    return 0;
}

static int
3052
qemuProcessInitPCIAddresses(virQEMUDriverPtr driver,
3053 3054
                            virDomainObjPtr vm,
                            int asyncJob)
3055 3056 3057
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int naddrs;
3058
    int ret = -1;
3059 3060
    qemuMonitorPCIAddress *addrs = NULL;

3061 3062
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        return -1;
3063 3064
    naddrs = qemuMonitorGetAllPCIAddresses(priv->mon,
                                           &addrs);
3065 3066
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        goto cleanup;
3067

3068 3069
    if (naddrs > 0)
        ret = qemuProcessDetectPCIAddresses(vm, addrs, naddrs);
3070

3071
 cleanup:
3072 3073 3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100
    VIR_FREE(addrs);

    return ret;
}


static int
qemuProcessPrepareChardevDevice(virDomainDefPtr def ATTRIBUTE_UNUSED,
                                virDomainChrDefPtr dev,
                                void *opaque ATTRIBUTE_UNUSED)
{
    int fd;
    if (dev->source.type != VIR_DOMAIN_CHR_TYPE_FILE)
        return 0;

    if ((fd = open(dev->source.data.file.path,
                   O_CREAT | O_APPEND, S_IRUSR|S_IWUSR)) < 0) {
        virReportSystemError(errno,
                             _("Unable to pre-create chardev file '%s'"),
                             dev->source.data.file.path);
        return -1;
    }

    VIR_FORCE_CLOSE(fd);

    return 0;
}


3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113
static int
qemuProcessCleanupChardevDevice(virDomainDefPtr def ATTRIBUTE_UNUSED,
                                virDomainChrDefPtr dev,
                                void *opaque ATTRIBUTE_UNUSED)
{
    if (dev->source.type == VIR_DOMAIN_CHR_TYPE_UNIX &&
        dev->source.data.nix.listen)
        unlink(dev->source.data.nix.path);

    return 0;
}


3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178
/**
 * Loads and update video memory size for video devices according to QEMU
 * process as the QEMU will silently update the values that we pass to QEMU
 * through command line.  We need to load these updated values and store them
 * into the status XML.
 *
 * We will fail if for some reason the values cannot be loaded from QEMU because
 * its mandatory to get the correct video memory size to status XML to not break
 * migration.
 */
static int
qemuProcessUpdateVideoRamSize(virQEMUDriverPtr driver,
                              virDomainObjPtr vm,
                              int asyncJob)
{
    int ret = -1;
    ssize_t i;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainVideoDefPtr video = NULL;
    virQEMUDriverConfigPtr cfg = NULL;

    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        return -1;

    for (i = 0; i < vm->def->nvideos; i++) {
        video = vm->def->videos[i];

        switch (video->type) {
        case VIR_DOMAIN_VIDEO_TYPE_VGA:
            if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_VGA_VGAMEM)) {
                if (qemuMonitorUpdateVideoMemorySize(priv->mon, video, "VGA") < 0)
                    goto error;
            }
            break;
        case VIR_DOMAIN_VIDEO_TYPE_QXL:
            if (i == 0) {
                if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_QXL_VGA_VGAMEM)) {
                    if (qemuMonitorUpdateVideoMemorySize(priv->mon, video,
                                                         "qxl-vga") < 0)
                        goto error;
                }
            } else {
                if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_QXL_VGAMEM)) {
                    if (qemuMonitorUpdateVideoMemorySize(priv->mon, video,
                                                         "qxl") < 0)
                        goto error;
                }
            }
            break;
        case VIR_DOMAIN_VIDEO_TYPE_VMVGA:
            if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_VMWARE_SVGA_VGAMEM)) {
                if (qemuMonitorUpdateVideoMemorySize(priv->mon, video,
                                                     "vmware-svga") < 0)
                    goto error;
            }
            break;
        case VIR_DOMAIN_VIDEO_TYPE_CIRRUS:
        case VIR_DOMAIN_VIDEO_TYPE_XEN:
        case VIR_DOMAIN_VIDEO_TYPE_VBOX:
        case VIR_DOMAIN_VIDEO_TYPE_LAST:
            break;
        }

    }

3179 3180
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        return -1;
3181 3182 3183 3184 3185 3186 3187 3188

    cfg = virQEMUDriverGetConfig(driver);
    ret = virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm);
    virObjectUnref(cfg);

    return ret;

 error:
3189
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
3190 3191 3192 3193
    return -1;
}


3194 3195 3196
struct qemuProcessHookData {
    virConnectPtr conn;
    virDomainObjPtr vm;
3197
    virQEMUDriverPtr driver;
3198
    virQEMUDriverConfigPtr cfg;
3199 3200 3201 3202 3203
};

static int qemuProcessHook(void *data)
{
    struct qemuProcessHookData *h = data;
3204
    qemuDomainObjPrivatePtr priv = h->vm->privateData;
3205
    int ret = -1;
3206
    int fd;
3207 3208 3209
    virBitmapPtr nodeset = NULL;
    virDomainNumatuneMemMode mode;

3210 3211 3212 3213 3214
    /* This method cannot use any mutexes, which are not
     * protected across fork()
     */

    virSecurityManagerPostFork(h->driver->securityManager);
3215 3216 3217 3218 3219

    /* Some later calls want pid present */
    h->vm->pid = getpid();

    VIR_DEBUG("Obtaining domain lock");
3220 3221 3222 3223 3224 3225 3226
    /*
     * Since we're going to leak the returned FD to QEMU,
     * we need to make sure it gets a sensible label.
     * This mildly sucks, because there could be other
     * sockets the lock driver opens that we don't want
     * labelled. So far we're ok though.
     */
3227
    if (virSecurityManagerSetSocketLabel(h->driver->securityManager, h->vm->def) < 0)
3228
        goto cleanup;
3229
    if (virDomainLockProcessStart(h->driver->lockManager,
3230
                                  h->cfg->uri,
3231
                                  h->vm,
J
Ján Tomko 已提交
3232
                                  /* QEMU is always paused initially */
3233 3234
                                  true,
                                  &fd) < 0)
3235
        goto cleanup;
3236
    if (virSecurityManagerClearSocketLabel(h->driver->securityManager, h->vm->def) < 0)
3237
        goto cleanup;
3238

3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249
    if (virDomainNumatuneGetMode(h->vm->def->numa, -1, &mode) == 0) {
        if (mode == VIR_DOMAIN_NUMATUNE_MEM_STRICT &&
            h->cfg->cgroupControllers & (1 << VIR_CGROUP_CONTROLLER_CPUSET) &&
            virCgroupControllerAvailable(VIR_CGROUP_CONTROLLER_CPUSET)) {
            /* Use virNuma* API iff necessary. Once set and child is exec()-ed,
             * there's no way for us to change it. Rely on cgroups (if available
             * and enabled in the config) rather than virNuma*. */
            VIR_DEBUG("Relying on CGroups for memory binding");
        } else {
            nodeset = virDomainNumatuneGetNodeset(h->vm->def->numa,
                                                  priv->autoNodeset, -1);
3250

3251 3252 3253
            if (virNumaSetupMemoryPolicy(mode, nodeset) < 0)
                goto cleanup;
        }
3254
    }
3255

3256 3257
    ret = 0;

3258
 cleanup:
3259
    virObjectUnref(h->cfg);
3260 3261
    VIR_DEBUG("Hook complete ret=%d", ret);
    return ret;
3262 3263 3264
}

int
3265
qemuProcessPrepareMonitorChr(virQEMUDriverConfigPtr cfg,
3266 3267 3268 3269 3270 3271
                             virDomainChrSourceDefPtr monConfig,
                             const char *vm)
{
    monConfig->type = VIR_DOMAIN_CHR_TYPE_UNIX;
    monConfig->data.nix.listen = true;

3272 3273 3274 3275
    if (virAsprintf(&monConfig->data.nix.path, "%s/%s.monitor",
                    cfg->libDir, vm) < 0)
        return -1;
    return 0;
3276 3277 3278
}


3279
/*
3280 3281
 * Precondition: vm must be locked, and a job must be active.
 * This method will call {Enter,Exit}Monitor
3282
 */
E
Eric Blake 已提交
3283
int
3284
qemuProcessStartCPUs(virQEMUDriverPtr driver, virDomainObjPtr vm,
3285
                     virConnectPtr conn, virDomainRunningReason reason,
3286
                     qemuDomainAsyncJob asyncJob)
3287
{
3288
    int ret = -1;
3289
    qemuDomainObjPrivatePtr priv = vm->privateData;
3290
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
3291

3292
    /* Bring up netdevs before starting CPUs */
3293
    if (qemuInterfaceStartDevices(vm->def) < 0)
3294 3295
       goto cleanup;

3296
    VIR_DEBUG("Using lock state '%s'", NULLSTR(priv->lockState));
3297
    if (virDomainLockProcessResume(driver->lockManager, cfg->uri,
3298
                                   vm, priv->lockState) < 0) {
3299 3300 3301 3302
        /* Don't free priv->lockState on error, because we need
         * to make sure we have state still present if the user
         * tries to resume again
         */
3303
        goto cleanup;
3304 3305 3306
    }
    VIR_FREE(priv->lockState);

3307 3308
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto release;
J
Jiri Denemark 已提交
3309

3310
    ret = qemuMonitorStartCPUs(priv->mon, conn);
3311 3312
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
3313 3314 3315 3316 3317

    if (ret < 0)
        goto release;

    virDomainObjSetState(vm, VIR_DOMAIN_RUNNING, reason);
3318

3319
 cleanup:
3320
    virObjectUnref(cfg);
3321
    return ret;
3322 3323 3324 3325 3326 3327

 release:
    if (virDomainLockProcessPause(driver->lockManager, vm, &priv->lockState) < 0)
        VIR_WARN("Unable to release lease on %s", vm->def->name);
    VIR_DEBUG("Preserving lock state '%s'", NULLSTR(priv->lockState));
    goto cleanup;
3328 3329 3330
}


3331 3332
int qemuProcessStopCPUs(virQEMUDriverPtr driver,
                        virDomainObjPtr vm,
3333
                        virDomainPausedReason reason,
3334
                        qemuDomainAsyncJob asyncJob)
3335
{
3336
    int ret = -1;
3337 3338
    qemuDomainObjPrivatePtr priv = vm->privateData;

3339
    VIR_FREE(priv->lockState);
J
Jiri Denemark 已提交
3340

3341 3342
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto cleanup;
J
Jiri Denemark 已提交
3343

3344
    ret = qemuMonitorStopCPUs(priv->mon);
3345 3346
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        ret = -1;
3347 3348 3349 3350

    if (ret < 0)
        goto cleanup;

3351 3352 3353
    /* de-activate netdevs after stopping CPUs */
    ignore_value(qemuInterfaceStopDevices(vm->def));

3354 3355 3356
    if (priv->job.current)
        ignore_value(virTimeMillisNow(&priv->job.current->stopped));

3357 3358 3359 3360
    virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, reason);
    if (virDomainLockProcessPause(driver->lockManager, vm, &priv->lockState) < 0)
        VIR_WARN("Unable to release lease on %s", vm->def->name);
    VIR_DEBUG("Preserving lock state '%s'", NULLSTR(priv->lockState));
J
Jiri Denemark 已提交
3361

3362
 cleanup:
3363 3364 3365 3366 3367
    return ret;
}



3368 3369 3370
static int
qemuProcessNotifyNets(virDomainDefPtr def)
{
3371
    size_t i;
3372

3373 3374
    for (i = 0; i < def->nnets; i++) {
        virDomainNetDefPtr net = def->nets[i];
3375
        if (networkNotifyActualDevice(def, net) < 0)
3376 3377 3378 3379 3380
            return -1;
    }
    return 0;
}

3381
static int
3382
qemuProcessFiltersInstantiate(virDomainDefPtr def)
3383
{
3384
    size_t i;
3385

3386
    for (i = 0; i < def->nnets; i++) {
3387 3388
        virDomainNetDefPtr net = def->nets[i];
        if ((net->filter) && (net->ifname)) {
3389
            if (virDomainConfNWFilterInstantiate(def->uuid, net) < 0)
J
Ján Tomko 已提交
3390
                return 1;
3391 3392 3393
        }
    }

J
Ján Tomko 已提交
3394
    return 0;
3395 3396
}

3397
static int
3398
qemuProcessUpdateState(virQEMUDriverPtr driver, virDomainObjPtr vm)
3399 3400 3401
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainState state;
3402
    virDomainPausedReason reason;
3403
    virDomainState newState = VIR_DOMAIN_NOSTATE;
3404
    int oldReason;
3405
    int newReason;
3406
    bool running;
3407
    char *msg = NULL;
3408 3409
    int ret;

3410
    qemuDomainObjEnterMonitor(driver, vm);
3411
    ret = qemuMonitorGetStatus(priv->mon, &running, &reason);
3412 3413
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
        return -1;
3414

3415
    if (ret < 0)
3416 3417
        return -1;

3418
    state = virDomainObjGetState(vm, &oldReason);
3419

3420 3421 3422 3423 3424 3425 3426 3427
    if (running &&
        (state == VIR_DOMAIN_SHUTOFF ||
         (state == VIR_DOMAIN_PAUSED &&
          oldReason == VIR_DOMAIN_PAUSED_STARTING_UP))) {
        newState = VIR_DOMAIN_RUNNING;
        newReason = VIR_DOMAIN_RUNNING_BOOTED;
        ignore_value(VIR_STRDUP_QUIET(msg, "finished booting"));
    } else if (state == VIR_DOMAIN_PAUSED && running) {
3428 3429
        newState = VIR_DOMAIN_RUNNING;
        newReason = VIR_DOMAIN_RUNNING_UNPAUSED;
3430
        ignore_value(VIR_STRDUP_QUIET(msg, "was unpaused"));
3431
    } else if (state == VIR_DOMAIN_RUNNING && !running) {
3432 3433 3434
        if (reason == VIR_DOMAIN_PAUSED_SHUTTING_DOWN) {
            newState = VIR_DOMAIN_SHUTDOWN;
            newReason = VIR_DOMAIN_SHUTDOWN_UNKNOWN;
3435
            ignore_value(VIR_STRDUP_QUIET(msg, "shutdown"));
3436
        } else if (reason == VIR_DOMAIN_PAUSED_CRASHED) {
3437 3438 3439
            newState = VIR_DOMAIN_CRASHED;
            newReason = VIR_DOMAIN_CRASHED_PANICKED;
            ignore_value(VIR_STRDUP_QUIET(msg, "crashed"));
3440 3441 3442
        } else {
            newState = VIR_DOMAIN_PAUSED;
            newReason = reason;
S
Stefan Berger 已提交
3443 3444
            ignore_value(virAsprintf(&msg, "was paused (%s)",
                                 virDomainPausedReasonTypeToString(reason)));
3445 3446 3447 3448 3449 3450 3451
        }
    }

    if (newState != VIR_DOMAIN_NOSTATE) {
        VIR_DEBUG("Domain %s %s while its monitor was disconnected;"
                  " changing state to %s (%s)",
                  vm->def->name,
3452
                  NULLSTR(msg),
3453 3454 3455 3456
                  virDomainStateTypeToString(newState),
                  virDomainStateReasonToString(newState, newReason));
        VIR_FREE(msg);
        virDomainObjSetState(vm, newState, newReason);
3457 3458 3459 3460 3461
    }

    return 0;
}

3462
static int
3463
qemuProcessRecoverMigration(virQEMUDriverPtr driver,
3464 3465
                            virDomainObjPtr vm,
                            virConnectPtr conn,
3466 3467
                            qemuDomainAsyncJob job,
                            qemuMigrationJobPhase phase,
3468 3469 3470 3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493
                            virDomainState state,
                            int reason)
{
    if (job == QEMU_ASYNC_JOB_MIGRATION_IN) {
        switch (phase) {
        case QEMU_MIGRATION_PHASE_NONE:
        case QEMU_MIGRATION_PHASE_PERFORM2:
        case QEMU_MIGRATION_PHASE_BEGIN3:
        case QEMU_MIGRATION_PHASE_PERFORM3:
        case QEMU_MIGRATION_PHASE_PERFORM3_DONE:
        case QEMU_MIGRATION_PHASE_CONFIRM3_CANCELLED:
        case QEMU_MIGRATION_PHASE_CONFIRM3:
        case QEMU_MIGRATION_PHASE_LAST:
            break;

        case QEMU_MIGRATION_PHASE_PREPARE:
            VIR_DEBUG("Killing unfinished incoming migration for domain %s",
                      vm->def->name);
            return -1;

        case QEMU_MIGRATION_PHASE_FINISH2:
            /* source domain is already killed so let's just resume the domain
             * and hope we are all set */
            VIR_DEBUG("Incoming migration finished, resuming domain %s",
                      vm->def->name);
            if (qemuProcessStartCPUs(driver, vm, conn,
3494 3495
                                     VIR_DOMAIN_RUNNING_UNPAUSED,
                                     QEMU_ASYNC_JOB_NONE) < 0) {
3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521 3522 3523
                VIR_WARN("Could not resume domain %s", vm->def->name);
            }
            break;

        case QEMU_MIGRATION_PHASE_FINISH3:
            /* migration finished, we started resuming the domain but didn't
             * confirm success or failure yet; killing it seems safest */
            VIR_DEBUG("Killing migrated domain %s", vm->def->name);
            return -1;
        }
    } else if (job == QEMU_ASYNC_JOB_MIGRATION_OUT) {
        switch (phase) {
        case QEMU_MIGRATION_PHASE_NONE:
        case QEMU_MIGRATION_PHASE_PREPARE:
        case QEMU_MIGRATION_PHASE_FINISH2:
        case QEMU_MIGRATION_PHASE_FINISH3:
        case QEMU_MIGRATION_PHASE_LAST:
            break;

        case QEMU_MIGRATION_PHASE_BEGIN3:
            /* nothing happen so far, just forget we were about to migrate the
             * domain */
            break;

        case QEMU_MIGRATION_PHASE_PERFORM2:
        case QEMU_MIGRATION_PHASE_PERFORM3:
            /* migration is still in progress, let's cancel it and resume the
             * domain */
3524
            if (qemuMigrationCancel(driver, vm) < 0)
3525
                return -1;
3526 3527 3528 3529 3530 3531
            /* resume the domain but only if it was paused as a result of
             * migration */
            if (state == VIR_DOMAIN_PAUSED &&
                (reason == VIR_DOMAIN_PAUSED_MIGRATION ||
                 reason == VIR_DOMAIN_PAUSED_UNKNOWN)) {
                if (qemuProcessStartCPUs(driver, vm, conn,
3532 3533
                                         VIR_DOMAIN_RUNNING_UNPAUSED,
                                         QEMU_ASYNC_JOB_NONE) < 0) {
3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552
                    VIR_WARN("Could not resume domain %s", vm->def->name);
                }
            }
            break;

        case QEMU_MIGRATION_PHASE_PERFORM3_DONE:
            /* migration finished but we didn't have a chance to get the result
             * of Finish3 step; third party needs to check what to do next
             */
            break;

        case QEMU_MIGRATION_PHASE_CONFIRM3_CANCELLED:
            /* Finish3 failed, we need to resume the domain */
            VIR_DEBUG("Resuming domain %s after failed migration",
                      vm->def->name);
            if (state == VIR_DOMAIN_PAUSED &&
                (reason == VIR_DOMAIN_PAUSED_MIGRATION ||
                 reason == VIR_DOMAIN_PAUSED_UNKNOWN)) {
                if (qemuProcessStartCPUs(driver, vm, conn,
3553 3554
                                         VIR_DOMAIN_RUNNING_UNPAUSED,
                                         QEMU_ASYNC_JOB_NONE) < 0) {
3555 3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568
                    VIR_WARN("Could not resume domain %s", vm->def->name);
                }
            }
            break;

        case QEMU_MIGRATION_PHASE_CONFIRM3:
            /* migration completed, we need to kill the domain here */
            return -1;
        }
    }

    return 0;
}

3569
static int
3570
qemuProcessRecoverJob(virQEMUDriverPtr driver,
3571 3572 3573 3574
                      virDomainObjPtr vm,
                      virConnectPtr conn,
                      const struct qemuDomainJobObj *job)
{
3575
    qemuDomainObjPrivatePtr priv = vm->privateData;
3576 3577 3578 3579 3580 3581 3582 3583
    virDomainState state;
    int reason;

    state = virDomainObjGetState(vm, &reason);

    switch (job->asyncJob) {
    case QEMU_ASYNC_JOB_MIGRATION_OUT:
    case QEMU_ASYNC_JOB_MIGRATION_IN:
3584 3585 3586
        if (qemuProcessRecoverMigration(driver, vm, conn, job->asyncJob,
                                        job->phase, state, reason) < 0)
            return -1;
3587 3588 3589 3590
        break;

    case QEMU_ASYNC_JOB_SAVE:
    case QEMU_ASYNC_JOB_DUMP:
3591
    case QEMU_ASYNC_JOB_SNAPSHOT:
3592
        qemuDomainObjEnterMonitor(driver, vm);
3593
        ignore_value(qemuMonitorMigrateCancel(priv->mon));
3594 3595
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            return -1;
3596
        /* resume the domain but only if it was paused as a result of
3597 3598 3599 3600 3601 3602 3603 3604 3605 3606 3607 3608 3609 3610 3611 3612
         * running a migration-to-file operation.  Although we are
         * recovering an async job, this function is run at startup
         * and must resume things using sync monitor connections.  */
         if (state == VIR_DOMAIN_PAUSED &&
             ((job->asyncJob == QEMU_ASYNC_JOB_DUMP &&
               reason == VIR_DOMAIN_PAUSED_DUMP) ||
              (job->asyncJob == QEMU_ASYNC_JOB_SAVE &&
               reason == VIR_DOMAIN_PAUSED_SAVE) ||
              (job->asyncJob == QEMU_ASYNC_JOB_SNAPSHOT &&
               reason == VIR_DOMAIN_PAUSED_SNAPSHOT) ||
              reason == VIR_DOMAIN_PAUSED_UNKNOWN)) {
             if (qemuProcessStartCPUs(driver, vm, conn,
                                      VIR_DOMAIN_RUNNING_UNPAUSED,
                                      QEMU_ASYNC_JOB_NONE) < 0) {
                 VIR_WARN("Could not resume domain '%s' after migration to file",
                          vm->def->name);
3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624
            }
        }
        break;

    case QEMU_ASYNC_JOB_NONE:
    case QEMU_ASYNC_JOB_LAST:
        break;
    }

    if (!virDomainObjIsActive(vm))
        return -1;

3625 3626 3627 3628
    /* In case any special handling is added for job type that has been ignored
     * before, QEMU_DOMAIN_TRACK_JOBS (from qemu_domain.h) needs to be updated
     * for the job to be properly tracked in domain state XML.
     */
3629 3630 3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648
    switch (job->active) {
    case QEMU_JOB_QUERY:
        /* harmless */
        break;

    case QEMU_JOB_DESTROY:
        VIR_DEBUG("Domain %s should have already been destroyed",
                  vm->def->name);
        return -1;

    case QEMU_JOB_SUSPEND:
        /* mostly harmless */
        break;

    case QEMU_JOB_MODIFY:
        /* XXX depending on the command we may be in an inconsistent state and
         * we should probably fall back to "monitor error" state and refuse to
         */
        break;

3649
    case QEMU_JOB_MIGRATION_OP:
3650
    case QEMU_JOB_ABORT:
3651 3652 3653 3654 3655 3656 3657 3658 3659 3660 3661
    case QEMU_JOB_ASYNC:
    case QEMU_JOB_ASYNC_NESTED:
        /* async job was already handled above */
    case QEMU_JOB_NONE:
    case QEMU_JOB_LAST:
        break;
    }

    return 0;
}

3662 3663 3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675 3676
static int
qemuProcessUpdateDevices(virQEMUDriverPtr driver,
                         virDomainObjPtr vm)
{
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virDomainDeviceDef dev;
    char **old;
    char **tmp;
    int ret = -1;

    if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE_DEL_EVENT))
        return 0;

    old = priv->qemuDevices;
    priv->qemuDevices = NULL;
3677
    if (qemuDomainUpdateDeviceList(driver, vm, QEMU_ASYNC_JOB_NONE) < 0)
3678 3679 3680 3681 3682
        goto cleanup;

    if ((tmp = old)) {
        while (*tmp) {
            if (!virStringArrayHasString(priv->qemuDevices, *tmp) &&
3683 3684 3685 3686
                virDomainDefFindDevice(vm->def, *tmp, &dev, false) == 0 &&
                qemuDomainRemoveDevice(driver, vm, &dev) < 0) {
                goto cleanup;
            }
3687 3688 3689 3690 3691
            tmp++;
        }
    }
    ret = 0;

3692
 cleanup:
3693 3694 3695 3696
    virStringFreeList(old);
    return ret;
}

3697 3698
struct qemuProcessReconnectData {
    virConnectPtr conn;
3699
    virQEMUDriverPtr driver;
3700
    virDomainObjPtr obj;
3701 3702 3703 3704
};
/*
 * Open an existing VM's monitor, re-detect VCPU threads
 * and re-reserve the security labels in use
S
Stefan Berger 已提交
3705 3706 3707 3708
 *
 * We own the virConnectPtr we are passed here - whoever started
 * this thread function has increased the reference counter to it
 * so that we now have to close it.
3709
 *
3710
 * This function also inherits a locked and ref'd domain object.
3711 3712 3713 3714 3715 3716 3717 3718 3719 3720 3721 3722
 *
 * This function needs to:
 * 1. Enter job
 * 1. just before monitor reconnect do lightweight MonitorEnter
 *    (increase VM refcount and unlock VM)
 * 2. reconnect to monitor
 * 3. do lightweight MonitorExit (lock VM)
 * 4. continue reconnect process
 * 5. EndJob
 *
 * We can't do normal MonitorEnter & MonitorExit because these two lock the
 * monitor lock, which does not exists in this early phase.
3723 3724
 */
static void
3725
qemuProcessReconnect(void *opaque)
3726 3727
{
    struct qemuProcessReconnectData *data = opaque;
3728
    virQEMUDriverPtr driver = data->driver;
3729
    virDomainObjPtr obj = data->obj;
3730 3731
    qemuDomainObjPrivatePtr priv;
    virConnectPtr conn = data->conn;
3732
    struct qemuDomainJobObj oldjob;
3733 3734
    int state;
    int reason;
3735
    virQEMUDriverConfigPtr cfg;
3736
    size_t i;
3737
    int ret;
3738
    unsigned int stopFlags = 0;
3739

3740 3741
    VIR_FREE(data);

3742
    qemuDomainObjRestoreJob(obj, &oldjob);
3743 3744
    if (oldjob.asyncJob == QEMU_ASYNC_JOB_MIGRATION_IN)
        stopFlags |= VIR_QEMU_PROCESS_STOP_MIGRATED;
3745

3746 3747 3748
    cfg = virQEMUDriverGetConfig(driver);
    priv = obj->privateData;

3749 3750 3751 3752 3753
    /* XXX If we ever gonna change pid file pattern, come up with
     * some intelligence here to deal with old paths. */
    if (!(priv->pidfile = virPidFileBuildPath(cfg->stateDir, obj->def->name)))
        goto killvm;

3754 3755 3756 3757
    if (qemuDomainObjBeginJob(driver, obj, QEMU_JOB_MODIFY) < 0)
        goto killvm;

    virNWFilterReadLockFilterUpdates();
3758 3759 3760 3761

    VIR_DEBUG("Reconnect monitor to %p '%s'", obj, obj->def->name);

    /* XXX check PID liveliness & EXE path */
3762
    if (qemuConnectMonitor(driver, obj, QEMU_ASYNC_JOB_NONE, -1) < 0)
3763 3764
        goto error;

D
Daniel P. Berrange 已提交
3765
    /* Failure to connect to agent shouldn't be fatal */
3766 3767 3768 3769
    if ((ret = qemuConnectAgent(driver, obj)) < 0) {
        if (ret == -2)
            goto error;

D
Daniel P. Berrange 已提交
3770 3771 3772 3773 3774 3775
        VIR_WARN("Cannot connect to QEMU guest agent for %s",
                 obj->def->name);
        virResetLastError();
        priv->agentError = true;
    }

3776
    if (qemuUpdateActivePCIHostdevs(driver, obj->def) < 0)
3777 3778
        goto error;

3779
    if (qemuUpdateActiveUSBHostdevs(driver, obj->def) < 0)
3780 3781
        goto error;

3782
    if (qemuUpdateActiveSCSIHostdevs(driver, obj->def) < 0)
3783 3784
        goto error;

3785
    if (qemuConnectCgroup(driver, obj) < 0)
3786 3787
        goto error;

3788
    /* XXX: Need to change as long as lock is introduced for
3789
     * qemu_driver->sharedDevices.
3790 3791
     */
    for (i = 0; i < obj->def->ndisks; i++) {
3792 3793
        virDomainDeviceDef dev;

3794
        if (virStorageTranslateDiskSourcePool(conn, obj->def->disks[i]) < 0)
3795
            goto error;
3796

3797 3798 3799 3800 3801 3802
        /* XXX we should be able to restore all data from XML in the future.
         * This should be the only place that calls qemuDomainDetermineDiskChain
         * with @report_broken == false to guarantee best-effort domain
         * reconnect */
        if (qemuDomainDetermineDiskChain(driver, obj, obj->def->disks[i],
                                         true, false) < 0)
3803 3804
            goto error;

3805 3806 3807
        dev.type = VIR_DOMAIN_DEVICE_DISK;
        dev.data.disk = obj->def->disks[i];
        if (qemuAddSharedDevice(driver, &dev, obj->def->name) < 0)
3808 3809 3810
            goto error;
    }

3811 3812 3813
    if (qemuProcessUpdateState(driver, obj) < 0)
        goto error;

3814
    state = virDomainObjGetState(obj, &reason);
3815 3816 3817
    if (state == VIR_DOMAIN_SHUTOFF ||
        (state == VIR_DOMAIN_PAUSED &&
         reason == VIR_DOMAIN_PAUSED_STARTING_UP)) {
3818 3819 3820 3821 3822
        VIR_DEBUG("Domain '%s' wasn't fully started yet, killing it",
                  obj->def->name);
        goto error;
    }

3823 3824 3825
    /* If upgrading from old libvirtd we won't have found any
     * caps in the domain status, so re-query them
     */
3826 3827
    if (!priv->qemuCaps &&
        !(priv->qemuCaps = virQEMUCapsCacheLookupCopy(driver->qemuCapsCache,
3828 3829
                                                      obj->def->emulator,
                                                      obj->def->os.machine)))
3830 3831
        goto error;

3832
    /* In case the domain shutdown while we were not running,
3833
     * we need to finish the shutdown process. And we need to do it after
3834
     * we have virQEMUCaps filled in.
3835
     */
3836 3837 3838 3839 3840 3841
    if (state == VIR_DOMAIN_SHUTDOWN ||
        (state == VIR_DOMAIN_PAUSED &&
         reason == VIR_DOMAIN_PAUSED_SHUTTING_DOWN)) {
        VIR_DEBUG("Finishing shutdown sequence for domain %s",
                  obj->def->name);
        qemuProcessShutdownOrReboot(driver, obj);
3842 3843
        qemuDomainObjEndJob(driver, obj);
        goto cleanup;
3844 3845
    }

3846 3847
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE))
        if ((qemuDomainAssignAddresses(obj->def, priv->qemuCaps, obj)) < 0)
3848
            goto error;
3849

3850 3851 3852 3853 3854 3855
    /* if domain requests security driver we haven't loaded, report error, but
     * do not kill the domain
     */
    ignore_value(virSecurityManagerCheckAllLabel(driver->securityManager,
                                                 obj->def));

3856
    if (virSecurityManagerReserveLabel(driver->securityManager, obj->def, obj->pid) < 0)
3857 3858
        goto error;

3859 3860 3861
    if (qemuProcessNotifyNets(obj->def) < 0)
        goto error;

3862
    if (qemuProcessFiltersInstantiate(obj->def))
3863 3864
        goto error;

3865
    if (qemuDomainCheckEjectableMedia(driver, obj, QEMU_ASYNC_JOB_NONE) < 0)
3866 3867
        goto error;

3868 3869 3870
    if (qemuProcessReconnectRefreshChannelVirtioState(driver, obj) < 0)
        goto error;

3871 3872 3873
    if (qemuProcessRefreshBalloonState(driver, obj, QEMU_ASYNC_JOB_NONE) < 0)
        goto error;

3874 3875 3876
    if (qemuProcessRecoverJob(driver, obj, conn, &oldjob) < 0)
        goto error;

3877 3878 3879
    if (qemuProcessUpdateDevices(driver, obj) < 0)
        goto error;

3880
    /* update domain state XML with possibly updated state in virDomainObj */
3881
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, obj) < 0)
3882 3883
        goto error;

3884 3885
    /* Run an hook to allow admins to do some magic */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
3886
        char *xml = qemuDomainDefFormatXML(driver, obj->def, 0);
3887 3888 3889 3890 3891 3892 3893 3894 3895 3896 3897 3898 3899 3900
        int hookret;

        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, obj->def->name,
                              VIR_HOOK_QEMU_OP_RECONNECT, VIR_HOOK_SUBOP_BEGIN,
                              NULL, xml, NULL);
        VIR_FREE(xml);

        /*
         * If the script raised an error abort the launch
         */
        if (hookret < 0)
            goto error;
    }

3901
    if (virAtomicIntInc(&driver->nactive) == 1 && driver->inhibitCallback)
3902 3903
        driver->inhibitCallback(true, driver->inhibitOpaque);

3904 3905
    qemuDomainObjEndJob(driver, obj);
    goto cleanup;
3906

3907
 error:
3908
    qemuDomainObjEndJob(driver, obj);
3909 3910 3911 3912 3913 3914 3915 3916 3917 3918 3919 3920 3921 3922 3923
 killvm:
    if (virDomainObjIsActive(obj)) {
        /* We can't get the monitor back, so must kill the VM
         * to remove danger of it ending up running twice if
         * user tries to start it again later
         */
        if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_NO_SHUTDOWN)) {
            /* If we couldn't get the monitor and qemu supports
             * no-shutdown, we can safely say that the domain
             * crashed ... */
            state = VIR_DOMAIN_SHUTOFF_CRASHED;
        } else {
            /* ... but if it doesn't we can't say what the state
             * really is and FAILED means "failed to start" */
            state = VIR_DOMAIN_SHUTOFF_UNKNOWN;
3924
        }
3925
        qemuProcessStop(driver, obj, state, stopFlags);
3926
    }
3927

3928 3929
    if (!obj->persistent)
        qemuDomainRemoveInactive(driver, obj);
3930

3931
 cleanup:
M
Michal Privoznik 已提交
3932
    virDomainObjEndAPI(&obj);
3933
    virObjectUnref(conn);
3934
    virObjectUnref(cfg);
3935
    virNWFilterUnlockFilterUpdates();
3936 3937
}

3938 3939
static int
qemuProcessReconnectHelper(virDomainObjPtr obj,
3940 3941 3942 3943 3944 3945
                           void *opaque)
{
    virThread thread;
    struct qemuProcessReconnectData *src = opaque;
    struct qemuProcessReconnectData *data;

3946
    /* If the VM was inactive, we don't need to reconnect */
3947 3948 3949
    if (!obj->pid)
        return 0;

3950
    if (VIR_ALLOC(data) < 0)
3951
        return -1;
3952 3953

    memcpy(data, src, sizeof(*data));
3954
    data->obj = obj;
3955

3956 3957
    /* this lock and reference will be eventually transferred to the thread
     * that handles the reconnect */
3958
    virObjectLock(obj);
3959
    virObjectRef(obj);
3960

3961 3962 3963
    /* Since we close the connection later on, we have to make sure that the
     * threads we start see a valid connection throughout their lifetime. We
     * simply increase the reference counter here.
S
Stefan Berger 已提交
3964
     */
3965
    virObjectRef(data->conn);
S
Stefan Berger 已提交
3966

3967
    if (virThreadCreate(&thread, false, qemuProcessReconnect, data) < 0) {
3968 3969 3970
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Could not create thread. QEMU initialization "
                         "might be incomplete"));
3971 3972 3973 3974
       /* We can't spawn a thread and thus connect to monitor. Kill qemu. */
        qemuProcessStop(src->driver, obj, VIR_DOMAIN_SHUTOFF_FAILED, 0);
        if (!obj->persistent)
            qemuDomainRemoveInactive(src->driver, obj);
3975

M
Michal Privoznik 已提交
3976
        virDomainObjEndAPI(&obj);
3977 3978 3979 3980
        virObjectUnref(data->conn);
        VIR_FREE(data);
        return -1;
    }
3981

3982
    return 0;
3983 3984 3985 3986 3987 3988 3989 3990 3991
}

/**
 * qemuProcessReconnectAll
 *
 * Try to re-open the resources for live VMs that we care
 * about.
 */
void
3992
qemuProcessReconnectAll(virConnectPtr conn, virQEMUDriverPtr driver)
3993
{
3994
    struct qemuProcessReconnectData data = {.conn = conn, .driver = driver};
3995
    virDomainObjListForEach(driver->domains, qemuProcessReconnectHelper, &data);
3996 3997
}

3998 3999 4000 4001 4002 4003 4004 4005 4006 4007 4008 4009 4010 4011 4012 4013 4014 4015 4016 4017 4018 4019 4020
static int
qemuProcessVNCAllocatePorts(virQEMUDriverPtr driver,
                            virDomainGraphicsDefPtr graphics)
{
    unsigned short port;

    if (graphics->data.vnc.socket)
        return 0;

    if (graphics->data.vnc.autoport) {
        if (virPortAllocatorAcquire(driver->remotePorts, &port) < 0)
            return -1;
        graphics->data.vnc.port = port;
    }

    if (graphics->data.vnc.websocket == -1) {
        if (virPortAllocatorAcquire(driver->webSocketPorts, &port) < 0)
            return -1;
        graphics->data.vnc.websocket = port;
    }

    return 0;
}
4021

4022
int
4023 4024
qemuProcessSPICEAllocatePorts(virQEMUDriverPtr driver,
                              virQEMUDriverConfigPtr cfg,
4025 4026
                              virDomainGraphicsDefPtr graphics,
                              bool allocate)
4027 4028 4029
{
    unsigned short port = 0;
    unsigned short tlsPort;
4030
    size_t i;
4031 4032 4033 4034 4035 4036 4037
    int defaultMode = graphics->data.spice.defaultMode;

    bool needTLSPort = false;
    bool needPort = false;

    if (graphics->data.spice.autoport) {
        /* check if tlsPort or port need allocation */
4038
        for (i = 0; i < VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_LAST; i++) {
4039 4040 4041 4042 4043 4044 4045 4046 4047 4048
            switch (graphics->data.spice.channels[i]) {
            case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_SECURE:
                needTLSPort = true;
                break;

            case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_INSECURE:
                needPort = true;
                break;

            case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_ANY:
4049
                /* default mode will be used */
4050 4051 4052
                break;
            }
        }
4053 4054 4055 4056 4057 4058 4059 4060 4061 4062 4063 4064 4065 4066 4067
        switch (defaultMode) {
        case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_SECURE:
            needTLSPort = true;
            break;

        case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_INSECURE:
            needPort = true;
            break;

        case VIR_DOMAIN_GRAPHICS_SPICE_CHANNEL_MODE_ANY:
            if (cfg->spiceTLS)
                needTLSPort = true;
            needPort = true;
            break;
        }
4068 4069
    }

4070 4071 4072 4073 4074 4075 4076 4077 4078 4079
    if (!allocate) {
        if (needPort || graphics->data.spice.port == -1)
            graphics->data.spice.port = 5901;

        if (needTLSPort || graphics->data.spice.tlsPort == -1)
            graphics->data.spice.tlsPort = 5902;

        return 0;
    }

4080
    if (needPort || graphics->data.spice.port == -1) {
4081
        if (virPortAllocatorAcquire(driver->remotePorts, &port) < 0)
4082
            goto error;
4083 4084 4085 4086

        graphics->data.spice.port = port;
    }

4087 4088
    if (needTLSPort || graphics->data.spice.tlsPort == -1) {
        if (!cfg->spiceTLS) {
4089 4090 4091 4092
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                           _("Auto allocation of spice TLS port requested "
                             "but spice TLS is disabled in qemu.conf"));
            goto error;
4093
        }
4094 4095 4096 4097 4098

        if (virPortAllocatorAcquire(driver->remotePorts, &tlsPort) < 0)
            goto error;

        graphics->data.spice.tlsPort = tlsPort;
4099 4100
    }

4101
    return 0;
4102

4103
 error:
4104
    virPortAllocatorRelease(driver->remotePorts, port);
4105
    return -1;
4106 4107 4108
}


4109 4110 4111 4112 4113 4114 4115 4116 4117 4118 4119 4120 4121 4122 4123 4124 4125 4126
static bool
qemuValidateCpuMax(virDomainDefPtr def, virQEMUCapsPtr qemuCaps)
{
    unsigned int maxCpus;

    maxCpus = virQEMUCapsGetMachineMaxCpus(qemuCaps, def->os.machine);
    if (!maxCpus)
        return true;

    if (def->maxvcpus > maxCpus) {
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
                       "%s", _("Maximum CPUs greater than specified machine type limit"));
        return false;
    }

    return true;
}

4127 4128

static bool
4129 4130 4131
qemuProcessVerifyGuestCPU(virQEMUDriverPtr driver,
                          virDomainObjPtr vm,
                          int asyncJob)
4132 4133 4134 4135 4136
{
    virDomainDefPtr def = vm->def;
    virArch arch = def->os.arch;
    virCPUDataPtr guestcpu = NULL;
    qemuDomainObjPrivatePtr priv = vm->privateData;
4137
    int rc;
4138
    bool ret = false;
J
Ján Tomko 已提交
4139
    size_t i;
4140

4141 4142 4143 4144 4145
    /* no features are passed to QEMU with -cpu host
     * so it makes no sense to verify them */
    if (def->cpu && def->cpu->mode == VIR_CPU_MODE_HOST_PASSTHROUGH)
        return true;

4146 4147 4148
    switch (arch) {
    case VIR_ARCH_I686:
    case VIR_ARCH_X86_64:
4149 4150
        if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
            return false;
4151
        rc = qemuMonitorGetGuestCPU(priv->mon, arch, &guestcpu);
4152 4153
        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            return false;
4154

4155 4156 4157 4158
        if (rc < 0) {
            if (rc == -2)
                break;

4159
            goto cleanup;
4160
        }
4161

J
Ján Tomko 已提交
4162
        if (def->features[VIR_DOMAIN_FEATURE_PVSPINLOCK] == VIR_TRISTATE_SWITCH_ON) {
4163 4164 4165 4166 4167 4168
            if (!cpuHasFeature(guestcpu, VIR_CPU_x86_KVM_PV_UNHALT)) {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                               _("host doesn't support paravirtual spinlocks"));
                goto cleanup;
            }
        }
J
Ján Tomko 已提交
4169

4170
        for (i = 0; def->cpu && i < def->cpu->nfeatures; i++) {
J
Ján Tomko 已提交
4171 4172 4173 4174 4175 4176 4177 4178 4179 4180 4181 4182
            virCPUFeatureDefPtr feature = &def->cpu->features[i];

            if (feature->policy != VIR_CPU_FEATURE_REQUIRE)
                continue;

            if (STREQ(feature->name, "invtsc") &&
                !cpuHasFeature(guestcpu, feature->name)) {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                               _("host doesn't support invariant TSC"));
                goto cleanup;
            }
        }
4183 4184 4185 4186 4187 4188 4189 4190
        break;

    default:
        break;
    }

    ret = true;

4191
 cleanup:
4192 4193 4194 4195 4196
    cpuDataFree(guestcpu);
    return ret;
}


4197 4198
static int
qemuPrepareNVRAM(virQEMUDriverConfigPtr cfg,
4199
                 virDomainObjPtr vm,
4200 4201 4202 4203 4204
                 bool migrated)
{
    int ret = -1;
    int srcFD = -1;
    int dstFD = -1;
4205
    virDomainLoaderDefPtr loader = vm->def->os.loader;
4206 4207 4208 4209 4210 4211 4212 4213 4214 4215 4216 4217 4218 4219 4220 4221 4222 4223 4224 4225 4226 4227 4228 4229 4230
    bool generated = false;
    bool created = false;

    /* Unless domain has RO loader of pflash type, we have
     * nothing to do here.  If the loader is RW then it's not
     * using split code and vars feature, so no nvram file needs
     * to be created. */
    if (!loader || loader->type != VIR_DOMAIN_LOADER_TYPE_PFLASH ||
        loader->readonly != VIR_TRISTATE_SWITCH_ON)
        return 0;

    /* If the nvram path is configured already, there's nothing
     * we need to do. Unless we are starting the destination side
     * of migration in which case nvram is configured in the
     * domain XML but the file doesn't exist yet. Moreover, after
     * the migration is completed, qemu will invoke a
     * synchronization write into the nvram file so we don't have
     * to take care about transmitting the real data on the other
     * side. */
    if (loader->nvram && !migrated)
        return 0;

    /* Autogenerate nvram path if needed.*/
    if (!loader->nvram) {
        if (virAsprintf(&loader->nvram,
4231 4232
                        "%s/%s_VARS.fd",
                        cfg->nvramDir, vm->def->name) < 0)
4233 4234 4235
            goto cleanup;

        generated = true;
4236

4237 4238
        if (vm->persistent &&
            virDomainSaveConfig(cfg->configDir, vm->def) < 0)
4239
            goto cleanup;
4240 4241 4242 4243 4244 4245 4246 4247 4248 4249 4250 4251 4252 4253 4254 4255 4256 4257 4258 4259 4260 4261 4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291 4292 4293 4294 4295 4296 4297 4298 4299 4300 4301 4302 4303 4304 4305 4306 4307 4308 4309 4310 4311 4312 4313 4314 4315 4316 4317 4318 4319 4320 4321 4322 4323 4324 4325 4326 4327 4328 4329
    }

    if (!virFileExists(loader->nvram)) {
        const char *master_nvram_path = loader->templt;
        ssize_t r;

        if (!loader->templt) {
            size_t i;
            for (i = 0; i < cfg->nloader; i++) {
                if (STREQ(cfg->loader[i], loader->path)) {
                    master_nvram_path = cfg->nvram[i];
                    break;
                }
            }
        }

        if (!master_nvram_path) {
            virReportError(VIR_ERR_OPERATION_FAILED,
                           _("unable to find any master var store for "
                             "loader: %s"), loader->path);
            goto cleanup;
        }

        if ((srcFD = virFileOpenAs(master_nvram_path, O_RDONLY,
                                   0, -1, -1, 0)) < 0) {
            virReportSystemError(-srcFD,
                                 _("Failed to open file '%s'"),
                                 master_nvram_path);
            goto cleanup;
        }
        if ((dstFD = virFileOpenAs(loader->nvram,
                                   O_WRONLY | O_CREAT | O_EXCL,
                                   S_IRUSR | S_IWUSR,
                                   cfg->user, cfg->group, 0)) < 0) {
            virReportSystemError(-dstFD,
                                 _("Failed to create file '%s'"),
                                 loader->nvram);
            goto cleanup;
        }
        created = true;

        do {
            char buf[1024];

            if ((r = saferead(srcFD, buf, sizeof(buf))) < 0) {
                virReportSystemError(errno,
                                     _("Unable to read from file '%s'"),
                                     master_nvram_path);
                goto cleanup;
            }

            if (safewrite(dstFD, buf, r) < 0) {
                virReportSystemError(errno,
                                     _("Unable to write to file '%s'"),
                                     loader->nvram);
                goto cleanup;
            }
        } while (r);

        if (VIR_CLOSE(srcFD) < 0) {
            virReportSystemError(errno,
                                 _("Unable to close file '%s'"),
                                 master_nvram_path);
            goto cleanup;
        }
        if (VIR_CLOSE(dstFD) < 0) {
            virReportSystemError(errno,
                                 _("Unable to close file '%s'"),
                                 loader->nvram);
            goto cleanup;
        }
    }

    ret = 0;
 cleanup:
    /* We successfully generated the nvram path, but failed to
     * copy the file content. Roll back. */
    if (ret < 0) {
        if (created)
            unlink(loader->nvram);
        if (generated)
            VIR_FREE(loader->nvram);
    }

    VIR_FORCE_CLOSE(srcFD);
    VIR_FORCE_CLOSE(dstFD);
    return ret;
}


4330 4331 4332 4333 4334 4335 4336 4337 4338 4339 4340 4341 4342 4343 4344 4345 4346 4347 4348 4349 4350 4351 4352 4353 4354 4355 4356 4357 4358 4359 4360 4361 4362 4363 4364 4365 4366 4367 4368
static void
qemuLogOperation(virDomainObjPtr vm,
                 const char *msg,
                 int logfd,
                 virCommandPtr cmd)
{
    char *timestamp;
    char *logline;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    int qemuVersion = virQEMUCapsGetVersion(priv->qemuCaps);
    const char *package = virQEMUCapsGetPackage(priv->qemuCaps);
    char ebuf[1024];

    if ((timestamp = virTimeStringNow()) == NULL)
        goto error;

    if (virAsprintf(&logline, "%s: %s %s, qemu version: %d.%d.%d%s\n",
                    timestamp, msg, VIR_LOG_VERSION_STRING,
                    (qemuVersion / 1000000) % 1000, (qemuVersion / 1000) % 1000, qemuVersion % 1000,
                    package ? package : "") < 0)
        goto error;

    if (safewrite(logfd, logline, strlen(logline)) < 0)
        goto error;

    if (cmd)
        virCommandWriteArgLog(cmd, logfd);

 cleanup:
    VIR_FREE(timestamp);
    VIR_FREE(logline);
    return;

 error:
    VIR_WARN("Unable to write banner to logfile: %s",
             virStrerror(errno, ebuf, sizeof(ebuf)));
    goto cleanup;
}

4369
int qemuProcessStart(virConnectPtr conn,
4370
                     virQEMUDriverPtr driver,
4371
                     virDomainObjPtr vm,
4372
                     int asyncJob,
4373 4374 4375
                     const char *migrateFrom,
                     int stdin_fd,
                     const char *stdin_path,
4376
                     virDomainSnapshotObjPtr snapshot,
4377
                     virNetDevVPortProfileOp vmop,
4378
                     unsigned int flags)
4379 4380 4381 4382 4383 4384 4385 4386
{
    int ret;
    off_t pos = -1;
    char ebuf[1024];
    int logfile = -1;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virCommandPtr cmd = NULL;
    struct qemuProcessHookData hookData;
4387
    size_t i;
4388
    bool rawio_set = false;
4389
    char *nodeset = NULL;
4390
    unsigned int stop_flags;
4391
    virQEMUDriverConfigPtr cfg;
4392
    virCapsPtr caps = NULL;
4393
    unsigned int hostdev_flags = 0;
4394 4395
    size_t nnicindexes = 0;
    int *nicindexes = NULL;
4396

4397 4398 4399 4400
    VIR_DEBUG("vm=%p name=%s id=%d asyncJob=%d migrateFrom=%s stdin_fd=%d "
              "stdin_path=%s snapshot=%p vmop=%d flags=0x%x",
              vm, vm->def->name, vm->def->id, asyncJob, NULLSTR(migrateFrom),
              stdin_fd, NULLSTR(stdin_path), snapshot, vmop, flags);
4401

4402 4403 4404 4405
    /* Okay, these are just internal flags,
     * but doesn't hurt to check */
    virCheckFlags(VIR_QEMU_PROCESS_START_COLD |
                  VIR_QEMU_PROCESS_START_PAUSED |
4406
                  VIR_QEMU_PROCESS_START_AUTODESTROY, -1);
4407

4408 4409
    cfg = virQEMUDriverGetConfig(driver);

4410 4411 4412 4413 4414
    /* From now on until domain security labeling is done:
     * if any operation fails and we goto cleanup, we must not
     * restore any security label as we would overwrite labels
     * we did not set. */
    stop_flags = VIR_QEMU_PROCESS_STOP_NO_RELABEL;
4415 4416 4417 4418
    /* If we fail while doing incoming migration, then we must not
     * relabel, as the source is still using the files.  */
    if (migrateFrom)
        stop_flags |= VIR_QEMU_PROCESS_STOP_MIGRATED;
4419

4420 4421 4422
    hookData.conn = conn;
    hookData.vm = vm;
    hookData.driver = driver;
4423 4424
    /* We don't increase cfg's reference counter here. */
    hookData.cfg = cfg;
4425

4426
    VIR_DEBUG("Beginning VM startup process");
4427 4428

    if (virDomainObjIsActive(vm)) {
4429 4430
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("VM is already active"));
4431
        virObjectUnref(cfg);
4432 4433 4434
        return -1;
    }

4435 4436 4437
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
        goto cleanup;

4438 4439 4440 4441
    /* Some things, paths, ... are generated here and we want them to persist.
     * Fill them in prior to setting the domain def as transient. */
    VIR_DEBUG("Generating paths");

4442
    if (qemuPrepareNVRAM(cfg, vm, migrateFrom) < 0)
4443 4444
        goto cleanup;

4445 4446 4447 4448
    /* Do this upfront, so any part of the startup process can add
     * runtime state to vm->def that won't be persisted. This let's us
     * report implicit runtime defaults in the XML, like vnc listen/socket
     */
4449
    VIR_DEBUG("Setting current domain def as transient");
4450
    if (virDomainObjSetDefTransient(caps, driver->xmlopt, vm, true) < 0)
4451 4452
        goto cleanup;

4453
    vm->def->id = qemuDriverAllocateID(driver);
4454
    qemuDomainSetFakeReboot(driver, vm, false);
4455
    virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, VIR_DOMAIN_PAUSED_STARTING_UP);
4456

4457
    if (virAtomicIntInc(&driver->nactive) == 1 && driver->inhibitCallback)
4458 4459
        driver->inhibitCallback(true, driver->inhibitOpaque);

4460
    /* Run an early hook to set-up missing devices */
4461
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
4462
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
4463 4464 4465
        int hookret;

        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
4466 4467
                              VIR_HOOK_QEMU_OP_PREPARE, VIR_HOOK_SUBOP_BEGIN,
                              NULL, xml, NULL);
4468 4469 4470 4471 4472 4473 4474 4475 4476
        VIR_FREE(xml);

        /*
         * If the script raised an error abort the launch
         */
        if (hookret < 0)
            goto cleanup;
    }

4477 4478 4479
    VIR_DEBUG("Determining emulator version");
    virObjectUnref(priv->qemuCaps);
    if (!(priv->qemuCaps = virQEMUCapsCacheLookupCopy(driver->qemuCapsCache,
4480 4481
                                                      vm->def->emulator,
                                                      vm->def->os.machine)))
4482 4483
        goto cleanup;

4484 4485 4486 4487 4488 4489 4490 4491
    /* network devices must be "prepared" before hostdevs, because
     * setting up a network device might create a new hostdev that
     * will need to be setup.
     */
    VIR_DEBUG("Preparing network devices");
    if (qemuNetworkPrepareDevices(vm->def) < 0)
       goto cleanup;

4492
    /* Must be run before security labelling */
4493
    VIR_DEBUG("Preparing host devices");
4494 4495
    if (!cfg->relaxedACS)
        hostdev_flags |= VIR_HOSTDEV_STRICT_ACS_CHECK;
4496 4497
    if (!migrateFrom)
        hostdev_flags |= VIR_HOSTDEV_COLD_BOOT;
4498
    if (qemuPrepareHostDevices(driver, vm->def, priv->qemuCaps,
4499
                               hostdev_flags) < 0)
4500 4501
        goto cleanup;

4502
    VIR_DEBUG("Preparing chr devices");
4503 4504 4505 4506 4507 4508
    if (virDomainChrDefForeach(vm->def,
                               true,
                               qemuProcessPrepareChardevDevice,
                               NULL) < 0)
        goto cleanup;

4509 4510 4511 4512
    VIR_DEBUG("Checking domain and device security labels");
    if (virSecurityManagerCheckAllLabel(driver->securityManager, vm->def) < 0)
        goto cleanup;

4513 4514
    /* If you are using a SecurityDriver with dynamic labelling,
       then generate a security label for isolation */
4515
    VIR_DEBUG("Generating domain security label (if required)");
4516
    if (virSecurityManagerGenLabel(driver->securityManager, vm->def) < 0) {
4517
        virDomainAuditSecurityLabel(vm, false);
4518 4519
        goto cleanup;
    }
4520
    virDomainAuditSecurityLabel(vm, true);
4521

4522
    if (vm->def->mem.nhugepages) {
4523 4524 4525 4526 4527 4528 4529 4530 4531 4532 4533 4534 4535 4536
        for (i = 0; i < cfg->nhugetlbfs; i++) {
            char *hugepagePath = qemuGetHugepagePath(&cfg->hugetlbfs[i]);

            if (!hugepagePath)
                goto cleanup;

            if (virSecurityManagerSetHugepages(driver->securityManager,
                                               vm->def, hugepagePath) < 0) {
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               "%s", _("Unable to set huge path in security driver"));
                VIR_FREE(hugepagePath);
                goto cleanup;
            }
            VIR_FREE(hugepagePath);
4537 4538 4539
        }
    }

4540 4541
    /* Ensure no historical cgroup for this VM is lying around bogus
     * settings */
4542
    VIR_DEBUG("Ensuring no historical cgroup is lying around");
4543
    qemuRemoveCgroup(driver, vm);
4544

4545 4546 4547 4548 4549 4550 4551 4552 4553 4554 4555 4556 4557 4558 4559 4560 4561 4562 4563 4564 4565 4566 4567 4568 4569 4570 4571 4572 4573 4574 4575 4576 4577 4578 4579
    for (i = 0; i < vm->def->ngraphics; ++i) {
        virDomainGraphicsDefPtr graphics = vm->def->graphics[i];
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC &&
            !graphics->data.vnc.autoport) {
            if (virPortAllocatorSetUsed(driver->remotePorts,
                                        graphics->data.vnc.port,
                                        true) < 0) {
                goto cleanup;
            }

            graphics->data.vnc.portReserved = true;

        } else if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE &&
                   !graphics->data.spice.autoport) {

            if (graphics->data.spice.port > 0) {
                if (virPortAllocatorSetUsed(driver->remotePorts,
                                            graphics->data.spice.port,
                                            true) < 0)
                    goto cleanup;

                graphics->data.spice.portReserved = true;
            }

            if (graphics->data.spice.tlsPort > 0) {
                if (virPortAllocatorSetUsed(driver->remotePorts,
                                            graphics->data.spice.tlsPort,
                                            true) < 0)
                    goto cleanup;

                graphics->data.spice.tlsPortReserved = true;
            }
        }
    }

4580
    for (i = 0; i < vm->def->ngraphics; ++i) {
4581
        virDomainGraphicsDefPtr graphics = vm->def->graphics[i];
4582 4583
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC) {
            if (qemuProcessVNCAllocatePorts(driver, graphics) < 0)
4584
                goto cleanup;
4585
        } else if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
4586
            if (qemuProcessSPICEAllocatePorts(driver, cfg, graphics, true) < 0)
4587
                goto cleanup;
4588
        }
4589

4590 4591
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC ||
            graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
4592
            if (graphics->nListens == 0) {
4593
                if (VIR_EXPAND_N(graphics->listens, graphics->nListens, 1) < 0)
4594 4595
                    goto cleanup;
                graphics->listens[0].type = VIR_DOMAIN_GRAPHICS_LISTEN_TYPE_ADDRESS;
4596 4597 4598
                if (VIR_STRDUP(graphics->listens[0].address,
                               graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC ?
                               cfg->vncListen : cfg->spiceListen) < 0) {
4599 4600 4601
                    VIR_SHRINK_N(graphics->listens, graphics->nListens, 1);
                    goto cleanup;
                }
4602
                graphics->listens[0].fromConfig = true;
4603 4604 4605 4606 4607
            } else if (graphics->nListens > 1) {
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                               _("QEMU does not support multiple listen "
                                 "addresses for one graphics device."));
                goto cleanup;
4608 4609
            }
        }
4610 4611
    }

4612
    if (virFileMakePath(cfg->logDir) < 0) {
4613 4614
        virReportSystemError(errno,
                             _("cannot create log directory %s"),
4615
                             cfg->logDir);
4616 4617 4618
        goto cleanup;
    }

4619
    VIR_DEBUG("Creating domain log file");
4620
    if ((logfile = qemuDomainCreateLog(driver, vm, false)) < 0)
4621 4622
        goto cleanup;

4623 4624
    if (vm->def->virtType == VIR_DOMAIN_VIRT_KVM) {
        VIR_DEBUG("Checking for KVM availability");
4625
        if (!virFileExists("/dev/kvm")) {
4626 4627 4628 4629
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                           _("Domain requires KVM, but it is not available. "
                             "Check that virtualization is enabled in the host BIOS, "
                             "and host configuration is setup to load the kvm modules."));
4630 4631 4632 4633
            goto cleanup;
        }
    }

4634 4635 4636
    if (!qemuValidateCpuMax(vm->def, priv->qemuCaps))
        goto cleanup;

4637
    if (qemuAssignDeviceAliases(vm->def, priv->qemuCaps) < 0)
4638 4639
        goto cleanup;

4640 4641 4642
    /* Get the advisory nodeset from numad if 'placement' of
     * either <vcpu> or <numatune> is 'auto'.
     */
4643
    if (virDomainDefNeedsPlacementAdvice(vm->def)) {
4644
        nodeset = virNumaGetAutoPlacementAdvice(vm->def->vcpus,
4645
                                                virDomainDefGetMemoryActual(vm->def));
4646 4647 4648 4649 4650
        if (!nodeset)
            goto cleanup;

        VIR_DEBUG("Nodeset returned from numad: %s", nodeset);

4651 4652 4653 4654 4655 4656
        if (virBitmapParse(nodeset, 0, &priv->autoNodeset,
                           VIR_DOMAIN_CPUMASK_LEN) < 0)
            goto cleanup;

        if (!(priv->autoCpuset = virCapabilitiesGetCpusForNodemask(caps,
                                                                   priv->autoNodeset)))
4657 4658 4659
            goto cleanup;
    }

4660 4661 4662
    if (virDomainDefCheckDuplicateDiskWWN(vm->def) < 0)
        goto cleanup;

4663 4664 4665 4666
    /* "volume" type disk's source must be translated before
     * cgroup and security setting.
     */
    for (i = 0; i < vm->def->ndisks; i++) {
4667
        if (virStorageTranslateDiskSourcePool(conn, vm->def->disks[i]) < 0)
4668 4669 4670
            goto cleanup;
    }

4671 4672 4673 4674
    if (qemuDomainCheckDiskPresence(driver, vm,
                                    flags & VIR_QEMU_PROCESS_START_COLD) < 0)
        goto cleanup;

4675 4676 4677 4678 4679 4680 4681
    if (vm->def->mem.min_guarantee) {
        virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                       _("Parameter 'min_guarantee' "
                         "not supported by QEMU."));
        goto cleanup;
    }

4682
    if (VIR_ALLOC(priv->monConfig) < 0)
4683 4684
        goto cleanup;

4685
    VIR_DEBUG("Preparing monitor state");
4686
    if (qemuProcessPrepareMonitorChr(cfg, priv->monConfig, vm->def->name) < 0)
4687 4688
        goto cleanup;

E
Eric Blake 已提交
4689
    priv->monJSON = virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_MONITOR_JSON);
4690 4691
    priv->monError = false;
    priv->monStart = 0;
4692 4693
    priv->gotShutdown = false;

4694
    VIR_FREE(priv->pidfile);
4695
    if (!(priv->pidfile = virPidFileBuildPath(cfg->stateDir, vm->def->name))) {
4696 4697
        virReportSystemError(errno,
                             "%s", _("Failed to build pidfile path."));
4698 4699 4700
        goto cleanup;
    }

4701 4702
    if (unlink(priv->pidfile) < 0 &&
        errno != ENOENT) {
4703
        virReportSystemError(errno,
4704 4705
                             _("Cannot remove stale PID file %s"),
                             priv->pidfile);
4706 4707 4708 4709 4710 4711 4712
        goto cleanup;
    }

    /*
     * Normally PCI addresses are assigned in the virDomainCreate
     * or virDomainDefine methods. We might still need to assign
     * some here to cope with the question of upgrades. Regardless
M
Martin Kletzander 已提交
4713
     * we also need to populate the PCI address set cache for later
4714 4715
     * use in hotplug
     */
4716
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE)) {
4717
        VIR_DEBUG("Assigning domain PCI addresses");
4718
        if ((qemuDomainAssignAddresses(vm->def, priv->qemuCaps, vm)) < 0)
4719
            goto cleanup;
4720 4721
    }

4722
    VIR_DEBUG("Building emulator command line");
4723
    if (!(cmd = qemuBuildCommandLine(conn, driver, vm->def, priv->monConfig,
E
Eric Blake 已提交
4724
                                     priv->monJSON, priv->qemuCaps,
4725
                                     migrateFrom, stdin_fd, snapshot, vmop,
4726
                                     &buildCommandLineCallbacks, false,
4727
                                     qemuCheckFips(),
4728
                                     priv->autoNodeset,
4729
                                     &nnicindexes, &nicindexes)))
4730 4731 4732 4733
        goto cleanup;

    /* now that we know it is about to start call the hook if present */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
4734
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
4735 4736 4737
        int hookret;

        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
4738 4739
                              VIR_HOOK_QEMU_OP_START, VIR_HOOK_SUBOP_BEGIN,
                              NULL, xml, NULL);
4740 4741 4742 4743 4744 4745 4746 4747 4748
        VIR_FREE(xml);

        /*
         * If the script raised an error abort the launch
         */
        if (hookret < 0)
            goto cleanup;
    }

4749
    qemuLogOperation(vm, "starting up", logfile, cmd);
4750

4751
    qemuDomainObjCheckTaint(driver, vm, logfile);
4752

4753 4754
    if ((pos = lseek(logfile, 0, SEEK_END)) < 0)
        VIR_WARN("Unable to seek to end of logfile: %s",
4755
                 virStrerror(errno, ebuf, sizeof(ebuf)));
4756 4757

    VIR_DEBUG("Clear emulator capabilities: %d",
4758 4759
              cfg->clearEmulatorCapabilities);
    if (cfg->clearEmulatorCapabilities)
4760 4761
        virCommandClearCaps(cmd);

4762 4763
    /* in case a certain disk is desirous of CAP_SYS_RAWIO, add this */
    for (i = 0; i < vm->def->ndisks; i++) {
4764
        virDomainDeviceDef dev;
4765 4766
        virDomainDiskDefPtr disk = vm->def->disks[i];

4767
        if (vm->def->disks[i]->rawio == VIR_TRISTATE_BOOL_YES) {
R
Roman Bogorodskiy 已提交
4768
#ifdef CAP_SYS_RAWIO
4769
            virCommandAllowCap(cmd, CAP_SYS_RAWIO);
4770
            rawio_set = true;
R
Roman Bogorodskiy 已提交
4771 4772 4773
#else
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                           _("Raw I/O is not supported on this platform"));
J
John Ferlan 已提交
4774
            goto cleanup;
R
Roman Bogorodskiy 已提交
4775
#endif
J
John Ferlan 已提交
4776
        }
4777

4778 4779 4780
        dev.type = VIR_DOMAIN_DEVICE_DISK;
        dev.data.disk = disk;
        if (qemuAddSharedDevice(driver, &dev, vm->def->name) < 0)
4781
            goto cleanup;
4782

4783
        if (qemuSetUnprivSGIO(&dev) < 0)
4784
            goto cleanup;
4785 4786
    }

4787 4788 4789 4790 4791 4792 4793 4794 4795 4796 4797 4798 4799 4800 4801 4802 4803 4804
    /* If rawio not already set, check hostdevs as well */
    if (!rawio_set) {
        for (i = 0; i < vm->def->nhostdevs; i++) {
            virDomainHostdevSubsysSCSIPtr scsisrc =
                &vm->def->hostdevs[i]->source.subsys.u.scsi;
            if (scsisrc->rawio == VIR_TRISTATE_BOOL_YES) {
#ifdef CAP_SYS_RAWIO
                virCommandAllowCap(cmd, CAP_SYS_RAWIO);
                break;
#else
                virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                               _("Raw I/O is not supported on this platform"));
                goto cleanup;
#endif
            }
        }
    }

4805
    virCommandSetPreExecHook(cmd, qemuProcessHook, &hookData);
4806 4807
    virCommandSetMaxProcesses(cmd, cfg->maxProcesses);
    virCommandSetMaxFiles(cmd, cfg->maxFiles);
4808
    virCommandSetUmask(cmd, 0x002);
4809

4810 4811 4812 4813 4814 4815
    VIR_DEBUG("Setting up security labelling");
    if (virSecurityManagerSetChildProcessLabel(driver->securityManager,
                                               vm->def, cmd) < 0) {
        goto cleanup;
    }

4816 4817 4818
    virCommandSetOutputFD(cmd, &logfile);
    virCommandSetErrorFD(cmd, &logfile);
    virCommandNonblockingFDs(cmd);
4819
    virCommandSetPidFile(cmd, priv->pidfile);
4820
    virCommandDaemonize(cmd);
4821
    virCommandRequireHandshake(cmd);
4822

4823 4824
    if (virSecurityManagerPreFork(driver->securityManager) < 0)
        goto cleanup;
4825
    ret = virCommandRun(cmd, NULL);
4826
    virSecurityManagerPostFork(driver->securityManager);
4827

E
Eric Blake 已提交
4828
    /* wait for qemu process to show up */
4829
    if (ret == 0) {
4830
        if (virPidFileReadPath(priv->pidfile, &vm->pid) < 0) {
4831 4832
            virReportError(VIR_ERR_INTERNAL_ERROR,
                           _("Domain %s didn't show up"), vm->def->name);
4833 4834
            ret = -1;
        }
4835 4836 4837 4838 4839
        VIR_DEBUG("QEMU vm=%p name=%s running with pid=%llu",
                  vm, vm->def->name, (unsigned long long)vm->pid);
    } else {
        VIR_DEBUG("QEMU vm=%p name=%s failed to spawn",
                  vm, vm->def->name);
4840 4841
    }

4842
    VIR_DEBUG("Writing early domain status to disk");
4843
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
4844 4845
        goto cleanup;

4846 4847
    VIR_DEBUG("Waiting for handshake from child");
    if (virCommandHandshakeWait(cmd) < 0) {
4848 4849
        /* Read errors from child that occurred between fork and exec. */
        qemuProcessReadChildErrors(driver, vm, pos);
4850 4851 4852
        goto cleanup;
    }

4853
    VIR_DEBUG("Setting up domain cgroup (if required)");
4854
    if (qemuSetupCgroup(driver, vm, nnicindexes, nicindexes) < 0)
4855 4856 4857 4858 4859
        goto cleanup;

    /* This must be done after cgroup placement to avoid resetting CPU
     * affinity */
    if (!vm->def->cputune.emulatorpin &&
4860
        qemuProcessInitCpuAffinity(vm) < 0)
4861 4862
        goto cleanup;

4863 4864
    VIR_DEBUG("Setting domain security labels");
    if (virSecurityManagerSetAllLabel(driver->securityManager,
4865
                                      vm->def, stdin_path) < 0)
4866 4867
        goto cleanup;

4868 4869 4870 4871 4872 4873
    /* Security manager labeled all devices, therefore
     * if any operation from now on fails and we goto cleanup,
     * where virSecurityManagerRestoreAllLabel() is called
     * (hidden under qemuProcessStop) we need to restore labels. */
    stop_flags &= ~VIR_QEMU_PROCESS_STOP_NO_RELABEL;

4874 4875 4876 4877 4878 4879 4880 4881 4882 4883 4884 4885 4886 4887
    if (stdin_fd != -1) {
        /* if there's an fd to migrate from, and it's a pipe, put the
         * proper security label on it
         */
        struct stat stdin_sb;

        VIR_DEBUG("setting security label on pipe used for migration");

        if (fstat(stdin_fd, &stdin_sb) < 0) {
            virReportSystemError(errno,
                                 _("cannot stat fd %d"), stdin_fd);
            goto cleanup;
        }
        if (S_ISFIFO(stdin_sb.st_mode) &&
4888
            virSecurityManagerSetImageFDLabel(driver->securityManager, vm->def, stdin_fd) < 0)
4889 4890 4891 4892
            goto cleanup;
    }

    VIR_DEBUG("Labelling done, completing handshake to child");
4893
    if (virCommandHandshakeNotify(cmd) < 0)
4894 4895 4896
        goto cleanup;
    VIR_DEBUG("Handshake complete, child running");

4897
    if (migrateFrom)
4898
        flags |= VIR_QEMU_PROCESS_START_PAUSED;
4899 4900 4901 4902 4903 4904 4905

    if (ret == -1) /* The VM failed to start; tear filters before taps */
        virDomainConfVMNWFilterTeardown(vm);

    if (ret == -1) /* The VM failed to start */
        goto cleanup;

4906
    VIR_DEBUG("Setting cgroup for emulator (if required)");
4907
    if (qemuSetupCgroupForEmulator(vm) < 0)
4908 4909 4910 4911 4912 4913
        goto cleanup;

    VIR_DEBUG("Setting affinity of emulator threads");
    if (qemuProcessSetEmulatorAffinity(vm) < 0)
        goto cleanup;

4914
    VIR_DEBUG("Waiting for monitor to show up");
4915
    if (qemuProcessWaitForMonitor(driver, vm, asyncJob, priv->qemuCaps, pos) < 0)
4916 4917
        goto cleanup;

D
Daniel P. Berrange 已提交
4918
    /* Failure to connect to agent shouldn't be fatal */
4919 4920 4921 4922
    if ((ret = qemuConnectAgent(driver, vm)) < 0) {
        if (ret == -2)
            goto cleanup;

D
Daniel P. Berrange 已提交
4923 4924 4925 4926 4927 4928
        VIR_WARN("Cannot connect to QEMU guest agent for %s",
                 vm->def->name);
        virResetLastError();
        priv->agentError = true;
    }

4929
    VIR_DEBUG("Detecting if required emulator features are present");
4930
    if (!qemuProcessVerifyGuestCPU(driver, vm, asyncJob))
4931 4932
        goto cleanup;

4933
    VIR_DEBUG("Setting up post-init cgroup restrictions");
4934
    if (qemuSetupCpusetMems(vm) < 0)
4935 4936
        goto cleanup;

4937
    VIR_DEBUG("Detecting VCPU PIDs");
4938
    if (qemuProcessDetectVcpuPIDs(driver, vm, asyncJob) < 0)
4939 4940
        goto cleanup;

4941 4942 4943 4944
    VIR_DEBUG("Detecting IOThread PIDs");
    if (qemuProcessDetectIOThreadPIDs(driver, vm, asyncJob) < 0)
        goto cleanup;

4945
    VIR_DEBUG("Setting cgroup for each VCPU (if required)");
4946
    if (qemuSetupCgroupForVcpu(vm) < 0)
4947 4948
        goto cleanup;

4949 4950 4951 4952
    VIR_DEBUG("Setting cgroup for each IOThread (if required)");
    if (qemuSetupCgroupForIOThreads(vm) < 0)
        goto cleanup;

4953
    VIR_DEBUG("Setting VCPU affinities");
4954
    if (qemuProcessSetVcpuAffinities(vm) < 0)
4955 4956
        goto cleanup;

4957 4958 4959 4960
    VIR_DEBUG("Setting affinity of IOThread threads");
    if (qemuProcessSetIOThreadsAffinity(vm) < 0)
        goto cleanup;

4961 4962 4963 4964
    VIR_DEBUG("Setting scheduler parameters");
    if (qemuProcessSetSchedulers(vm) < 0)
        goto cleanup;

4965
    VIR_DEBUG("Setting any required VM passwords");
4966
    if (qemuProcessInitPasswords(conn, driver, vm, asyncJob) < 0)
4967 4968 4969 4970
        goto cleanup;

    /* If we have -device, then addresses are assigned explicitly.
     * If not, then we have to detect dynamic ones here */
4971
    if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE)) {
4972
        VIR_DEBUG("Determining domain device PCI addresses");
4973
        if (qemuProcessInitPCIAddresses(driver, vm, asyncJob) < 0)
4974 4975 4976
            goto cleanup;
    }

4977 4978 4979 4980
    /* set default link states */
    /* qemu doesn't support setting this on the command line, so
     * enter the monitor */
    VIR_DEBUG("Setting network link states");
4981 4982
    if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
        goto cleanup;
4983 4984 4985
    if (qemuProcessSetLinkStates(vm) < 0)
        goto exit_monitor;
    if (qemuDomainObjExitMonitor(driver, vm))
4986 4987
        goto cleanup;

4988
    VIR_DEBUG("Fetching list of active devices");
4989
    if (qemuDomainUpdateDeviceList(driver, vm, asyncJob) < 0)
4990 4991
        goto cleanup;

4992 4993 4994 4995
    VIR_DEBUG("Updating info of memory devices");
    if (qemuDomainUpdateMemoryDeviceInfo(driver, vm, asyncJob) < 0)
        goto cleanup;

4996
    VIR_DEBUG("Setting initial memory amount");
4997 4998 4999 5000 5001 5002 5003 5004 5005 5006 5007 5008 5009 5010 5011 5012
    if (vm->def->memballoon &&
        vm->def->memballoon->model != VIR_DOMAIN_MEMBALLOON_MODEL_NONE) {
        unsigned long long balloon = vm->def->mem.cur_balloon;
        int period = vm->def->memballoon->period;

        if (qemuDomainObjEnterMonitorAsync(driver, vm, asyncJob) < 0)
            goto cleanup;

        if (period)
            qemuMonitorSetMemoryStatsPeriod(priv->mon, period);

        if (qemuMonitorSetBalloon(priv->mon, balloon) < 0)
            goto exit_monitor;

        if (qemuDomainObjExitMonitor(driver, vm) < 0)
            goto cleanup;
5013
    }
5014

5015
    /* Since CPUs were not started yet, the balloon could not return the memory
5016 5017
     * to the host and thus cur_balloon needs to be updated so that GetXMLdesc
     * and friends return the correct size in case they can't grab the job */
5018 5019
    if (!migrateFrom && !snapshot &&
        qemuProcessRefreshBalloonState(driver, vm, asyncJob) < 0)
5020
        goto cleanup;
5021

5022 5023 5024 5025
    VIR_DEBUG("Detecting actual memory size for video device");
    if (qemuProcessUpdateVideoRamSize(driver, vm, asyncJob) < 0)
        goto cleanup;

5026
    if (!(flags & VIR_QEMU_PROCESS_START_PAUSED)) {
5027
        VIR_DEBUG("Starting domain CPUs");
5028
        /* Allow the CPUS to start executing */
J
Jiri Denemark 已提交
5029
        if (qemuProcessStartCPUs(driver, vm, conn,
5030 5031
                                 VIR_DOMAIN_RUNNING_BOOTED,
                                 QEMU_ASYNC_JOB_NONE) < 0) {
5032
            if (virGetLastError() == NULL)
5033 5034
                virReportError(VIR_ERR_INTERNAL_ERROR,
                               "%s", _("resume operation failed"));
5035 5036
            goto cleanup;
        }
J
Jiri Denemark 已提交
5037 5038 5039 5040 5041
    } else {
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED,
                             migrateFrom ?
                             VIR_DOMAIN_PAUSED_MIGRATION :
                             VIR_DOMAIN_PAUSED_USER);
5042 5043
    }

5044
    if (flags & VIR_QEMU_PROCESS_START_AUTODESTROY &&
5045 5046 5047
        qemuProcessAutoDestroyAdd(driver, vm, conn) < 0)
        goto cleanup;

5048
    VIR_DEBUG("Writing domain status to disk");
5049
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
5050 5051
        goto cleanup;

5052 5053
    /* finally we can call the 'started' hook script if any */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
5054
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
5055 5056 5057 5058 5059 5060 5061 5062 5063 5064 5065 5066 5067 5068
        int hookret;

        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
                              VIR_HOOK_QEMU_OP_STARTED, VIR_HOOK_SUBOP_BEGIN,
                              NULL, xml, NULL);
        VIR_FREE(xml);

        /*
         * If the script raised an error abort the launch
         */
        if (hookret < 0)
            goto cleanup;
    }

5069 5070 5071 5072
    /* Keep watching qemu log for errors during incoming migration, otherwise
     * unset reporting errors from qemu log. */
    if (!migrateFrom)
        qemuMonitorSetDomainLog(priv->mon, -1);
5073

5074 5075
    virCommandFree(cmd);
    VIR_FORCE_CLOSE(logfile);
5076
    virObjectUnref(cfg);
5077
    virObjectUnref(caps);
5078
    VIR_FREE(nicindexes);
5079 5080 5081

    return 0;

5082
 cleanup:
5083 5084 5085
    /* We jump here if we failed to start the VM for any reason, or
     * if we failed to initialize the now running VM. kill it off and
     * pretend we never started it */
5086
    VIR_FREE(nodeset);
5087 5088
    virCommandFree(cmd);
    VIR_FORCE_CLOSE(logfile);
5089 5090
    if (priv->mon)
        qemuMonitorSetDomainLog(priv->mon, -1);
5091
    qemuProcessStop(driver, vm, VIR_DOMAIN_SHUTOFF_FAILED, stop_flags);
5092
    virObjectUnref(cfg);
5093
    virObjectUnref(caps);
5094
    VIR_FREE(nicindexes);
5095 5096

    return -1;
5097 5098 5099 5100

 exit_monitor:
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
    goto cleanup;
5101 5102 5103
}


5104
int
5105
qemuProcessKill(virDomainObjPtr vm, unsigned int flags)
5106
{
5107
    int ret;
5108

5109 5110 5111
    VIR_DEBUG("vm=%p name=%s pid=%llu flags=%x",
              vm, vm->def->name,
              (unsigned long long)vm->pid, flags);
5112

5113 5114 5115 5116 5117
    if (!(flags & VIR_QEMU_PROCESS_KILL_NOCHECK)) {
        if (!virDomainObjIsActive(vm)) {
            VIR_DEBUG("VM '%s' not active", vm->def->name);
            return 0;
        }
5118 5119
    }

5120
    if (flags & VIR_QEMU_PROCESS_KILL_NOWAIT) {
5121 5122 5123 5124 5125
        virProcessKill(vm->pid,
                       (flags & VIR_QEMU_PROCESS_KILL_FORCE) ?
                       SIGKILL : SIGTERM);
        return 0;
    }
5126

5127 5128
    ret = virProcessKillPainfully(vm->pid,
                                  !!(flags & VIR_QEMU_PROCESS_KILL_FORCE));
5129

5130
    return ret;
5131 5132 5133
}


5134
void qemuProcessStop(virQEMUDriverPtr driver,
5135
                     virDomainObjPtr vm,
5136 5137
                     virDomainShutoffReason reason,
                     unsigned int flags)
5138 5139 5140 5141 5142 5143
{
    int ret;
    int retries = 0;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    virErrorPtr orig_err;
    virDomainDefPtr def;
A
Ansis Atteka 已提交
5144
    virNetDevVPortProfilePtr vport = NULL;
5145
    size_t i;
5146 5147 5148
    int logfile = -1;
    char *timestamp;
    char ebuf[1024];
5149
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
5150

5151 5152 5153
    VIR_DEBUG("Shutting down vm=%p name=%s id=%d pid=%llu flags=%x",
              vm, vm->def->name, vm->def->id,
              (unsigned long long)vm->pid, flags);
5154 5155 5156

    if (!virDomainObjIsActive(vm)) {
        VIR_DEBUG("VM '%s' not active", vm->def->name);
5157
        virObjectUnref(cfg);
5158 5159 5160
        return;
    }

5161 5162 5163 5164
    /* This method is routinely used in clean up paths. Disable error
     * reporting so we don't squash a legit error. */
    orig_err = virSaveLastError();

5165
    /*
5166 5167
     * We may unlock the vm in qemuProcessKill(), and another thread
     * can lock the vm, and then call qemuProcessStop(). So we should
5168 5169 5170 5171
     * set vm->def->id to -1 here to avoid qemuProcessStop() to be called twice.
     */
    vm->def->id = -1;

5172
    if (virAtomicIntDecAndTest(&driver->nactive) && driver->inhibitCallback)
5173 5174
        driver->inhibitCallback(false, driver->inhibitOpaque);

5175 5176
    /* Wake up anything waiting on domain condition */
    virDomainObjBroadcast(vm);
5177

5178
    if ((logfile = qemuDomainCreateLog(driver, vm, true)) < 0) {
5179 5180 5181
        /* To not break the normal domain shutdown process, skip the
         * timestamp log writing if failed on opening log file. */
        VIR_WARN("Unable to open logfile: %s",
5182
                  virStrerror(errno, ebuf, sizeof(ebuf)));
5183
    } else {
5184
        if ((timestamp = virTimeStringNow()) != NULL) {
5185 5186 5187 5188
            if (safewrite(logfile, timestamp, strlen(timestamp)) < 0 ||
                safewrite(logfile, SHUTDOWN_POSTFIX,
                          strlen(SHUTDOWN_POSTFIX)) < 0) {
                VIR_WARN("Unable to write timestamp to logfile: %s",
5189
                         virStrerror(errno, ebuf, sizeof(ebuf)));
5190 5191 5192 5193 5194 5195 5196
            }

            VIR_FREE(timestamp);
        }

        if (VIR_CLOSE(logfile) < 0)
             VIR_WARN("Unable to close logfile: %s",
5197
                      virStrerror(errno, ebuf, sizeof(ebuf)));
5198 5199
    }

5200 5201 5202
    /* Clear network bandwidth */
    virDomainClearNetBandwidth(vm);

5203 5204
    virDomainConfVMNWFilterTeardown(vm);

5205
    if (cfg->macFilter) {
5206
        def = vm->def;
5207
        for (i = 0; i < def->nnets; i++) {
5208 5209 5210
            virDomainNetDefPtr net = def->nets[i];
            if (net->ifname == NULL)
                continue;
5211 5212 5213
            ignore_value(ebtablesRemoveForwardAllowIn(driver->ebtables,
                                                      net->ifname,
                                                      &net->mac));
5214 5215 5216
        }
    }

5217
    virPortAllocatorRelease(driver->migrationPorts, priv->nbdPort);
5218
    priv->nbdPort = 0;
5219

D
Daniel P. Berrange 已提交
5220 5221 5222 5223 5224 5225
    if (priv->agent) {
        qemuAgentClose(priv->agent);
        priv->agent = NULL;
        priv->agentError = false;
    }

5226
    if (priv->mon) {
5227
        qemuMonitorClose(priv->mon);
5228 5229
        priv->mon = NULL;
    }
5230 5231 5232 5233 5234 5235 5236 5237

    if (priv->monConfig) {
        if (priv->monConfig->type == VIR_DOMAIN_CHR_TYPE_UNIX)
            unlink(priv->monConfig->data.nix.path);
        virDomainChrSourceDefFree(priv->monConfig);
        priv->monConfig = NULL;
    }

5238 5239 5240 5241 5242 5243
    ignore_value(virDomainChrDefForeach(vm->def,
                                        false,
                                        qemuProcessCleanupChardevDevice,
                                        NULL));


5244
    /* shut it off for sure */
5245 5246 5247
    ignore_value(qemuProcessKill(vm,
                                 VIR_QEMU_PROCESS_KILL_FORCE|
                                 VIR_QEMU_PROCESS_KILL_NOCHECK));
5248

5249 5250
    qemuDomainCleanupRun(driver, vm);

5251
    /* Stop autodestroy in case guest is restarted */
5252
    qemuProcessAutoDestroyRemove(driver, vm);
5253

5254 5255
    /* now that we know it's stopped call the hook if present */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
5256
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
5257 5258 5259

        /* we can't stop the operation even if the script raised an error */
        virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
5260 5261
                    VIR_HOOK_QEMU_OP_STOPPED, VIR_HOOK_SUBOP_END,
                    NULL, xml, NULL);
5262 5263 5264
        VIR_FREE(xml);
    }

5265 5266 5267 5268
    /* Reset Security Labels unless caller don't want us to */
    if (!(flags & VIR_QEMU_PROCESS_STOP_NO_RELABEL))
        virSecurityManagerRestoreAllLabel(driver->securityManager,
                                          vm->def,
5269
                                          !!(flags & VIR_QEMU_PROCESS_STOP_MIGRATED));
5270
    virSecurityManagerReleaseLabel(driver->securityManager, vm->def);
5271

5272
    for (i = 0; i < vm->def->ndisks; i++) {
5273
        virDomainDeviceDef dev;
5274
        virDomainDiskDefPtr disk = vm->def->disks[i];
5275 5276 5277 5278

        dev.type = VIR_DOMAIN_DEVICE_DISK;
        dev.data.disk = disk;
        ignore_value(qemuRemoveSharedDevice(driver, &dev, vm->def->name));
5279 5280
    }

5281
    /* Clear out dynamically assigned labels */
5282
    for (i = 0; i < vm->def->nseclabels; i++) {
5283
        if (vm->def->seclabels[i]->type == VIR_DOMAIN_SECLABEL_DYNAMIC)
5284 5285
            VIR_FREE(vm->def->seclabels[i]->label);
        VIR_FREE(vm->def->seclabels[i]->imagelabel);
5286 5287
    }

5288 5289 5290
    virStringFreeList(priv->qemuDevices);
    priv->qemuDevices = NULL;

5291 5292 5293
    virDomainDefClearDeviceAliases(vm->def);
    if (!priv->persistentAddrs) {
        virDomainDefClearPCIAddresses(vm->def);
5294
        virDomainPCIAddressSetFree(priv->pciaddrs);
5295
        priv->pciaddrs = NULL;
5296
        virDomainDefClearCCWAddresses(vm->def);
J
Ján Tomko 已提交
5297
        virDomainCCWAddressSetFree(priv->ccwaddrs);
5298
        priv->ccwaddrs = NULL;
5299 5300
        virDomainVirtioSerialAddrSetFree(priv->vioserialaddrs);
        priv->vioserialaddrs = NULL;
5301 5302 5303 5304 5305 5306 5307
    }

    qemuDomainReAttachHostDevices(driver, vm->def);

    def = vm->def;
    for (i = 0; i < def->nnets; i++) {
        virDomainNetDefPtr net = def->nets[i];
5308 5309 5310 5311
        vport = virDomainNetGetActualVirtPortProfile(net);

        switch (virDomainNetGetActualType(net)) {
        case VIR_DOMAIN_NET_TYPE_DIRECT:
5312
            ignore_value(virNetDevMacVLanDeleteWithVPortProfile(
5313
                             net->ifname, &net->mac,
5314 5315
                             virDomainNetGetActualDirectDev(net),
                             virDomainNetGetActualDirectMode(net),
5316
                             virDomainNetGetActualVirtPortProfile(net),
5317
                             cfg->stateDir));
5318
            VIR_FREE(net->ifname);
5319 5320 5321 5322 5323
            break;
        case VIR_DOMAIN_NET_TYPE_BRIDGE:
        case VIR_DOMAIN_NET_TYPE_NETWORK:
#ifdef VIR_NETDEV_TAP_REQUIRE_MANUAL_CLEANUP
            if (!(vport && vport->virtPortType == VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH))
5324
                ignore_value(virNetDevTapDelete(net->ifname, net->backend.tap));
5325 5326
#endif
            break;
5327
        }
5328 5329 5330
        /* release the physical device (or any other resources used by
         * this interface in the network driver
         */
5331 5332 5333 5334 5335 5336 5337 5338 5339
        if (vport) {
            if (vport->virtPortType == VIR_NETDEV_VPORT_PROFILE_MIDONET) {
                ignore_value(virNetDevMidonetUnbindPort(vport));
            } else if (vport->virtPortType == VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH) {
                ignore_value(virNetDevOpenvswitchRemovePort(
                                 virDomainNetGetActualBridgeName(net),
                                 net->ifname));
            }
        }
A
Ansis Atteka 已提交
5340

5341 5342
        /* kick the device out of the hostdev list too */
        virDomainNetRemoveHostdev(def, net);
5343
        networkReleaseActualDevice(vm->def, net);
5344
    }
5345

5346
 retry:
5347
    if ((ret = qemuRemoveCgroup(driver, vm)) < 0) {
5348 5349 5350 5351 5352 5353 5354
        if (ret == -EBUSY && (retries++ < 5)) {
            usleep(200*1000);
            goto retry;
        }
        VIR_WARN("Failed to remove cgroup for %s",
                 vm->def->name);
    }
5355
    virCgroupFree(&priv->cgroup);
5356 5357 5358

    qemuProcessRemoveDomainStatus(driver, vm);

5359 5360
    /* Remove VNC and Spice ports from port reservation bitmap, but only if
       they were reserved by the driver (autoport=yes)
5361
    */
5362
    for (i = 0; i < vm->def->ngraphics; ++i) {
5363
        virDomainGraphicsDefPtr graphics = vm->def->graphics[i];
5364 5365
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_VNC) {
            if (graphics->data.vnc.autoport) {
5366 5367
                virPortAllocatorRelease(driver->remotePorts,
                                        graphics->data.vnc.port);
5368
            } else if (graphics->data.vnc.portReserved) {
5369 5370 5371 5372 5373
                virPortAllocatorSetUsed(driver->remotePorts,
                                        graphics->data.spice.port,
                                        false);
                graphics->data.vnc.portReserved = false;
            }
5374 5375
            virPortAllocatorRelease(driver->webSocketPorts,
                                    graphics->data.vnc.websocket);
5376
        }
5377 5378 5379 5380 5381 5382 5383 5384 5385 5386 5387 5388 5389 5390 5391 5392 5393 5394 5395 5396 5397
        if (graphics->type == VIR_DOMAIN_GRAPHICS_TYPE_SPICE) {
            if (graphics->data.spice.autoport) {
                virPortAllocatorRelease(driver->remotePorts,
                                        graphics->data.spice.port);
                virPortAllocatorRelease(driver->remotePorts,
                                        graphics->data.spice.tlsPort);
            } else {
                if (graphics->data.spice.portReserved) {
                    virPortAllocatorSetUsed(driver->remotePorts,
                                            graphics->data.spice.port,
                                            false);
                    graphics->data.spice.portReserved = false;
                }

                if (graphics->data.spice.tlsPortReserved) {
                    virPortAllocatorSetUsed(driver->remotePorts,
                                            graphics->data.spice.tlsPort,
                                            false);
                    graphics->data.spice.tlsPortReserved = false;
                }
            }
5398
        }
5399 5400
    }

5401
    vm->taint = 0;
5402
    vm->pid = -1;
J
Jiri Denemark 已提交
5403
    virDomainObjSetState(vm, VIR_DOMAIN_SHUTOFF, reason);
5404 5405
    VIR_FREE(priv->vcpupids);
    priv->nvcpupids = 0;
5406 5407
    for (i = 0; i < vm->def->niothreadids; i++)
        vm->def->iothreadids[i]->thread_id = 0;
5408 5409
    virObjectUnref(priv->qemuCaps);
    priv->qemuCaps = NULL;
5410
    VIR_FREE(priv->pidfile);
5411

5412
    /* The "release" hook cleans up additional resources */
5413
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
5414
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
5415 5416 5417

        /* we can't stop the operation even if the script raised an error */
        virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
5418 5419
                    VIR_HOOK_QEMU_OP_RELEASE, VIR_HOOK_SUBOP_END,
                    NULL, xml, NULL);
5420 5421 5422
        VIR_FREE(xml);
    }

5423 5424 5425 5426 5427 5428 5429 5430 5431 5432 5433
    if (vm->newDef) {
        virDomainDefFree(vm->def);
        vm->def = vm->newDef;
        vm->def->id = -1;
        vm->newDef = NULL;
    }

    if (orig_err) {
        virSetError(orig_err);
        virFreeError(orig_err);
    }
5434
    virObjectUnref(cfg);
5435
}
5436 5437


5438
int qemuProcessAttach(virConnectPtr conn ATTRIBUTE_UNUSED,
5439
                      virQEMUDriverPtr driver,
5440
                      virDomainObjPtr vm,
5441
                      pid_t pid,
5442 5443 5444 5445
                      const char *pidfile,
                      virDomainChrSourceDefPtr monConfig,
                      bool monJSON)
{
5446
    size_t i;
5447 5448 5449 5450 5451
    char ebuf[1024];
    int logfile = -1;
    char *timestamp;
    qemuDomainObjPrivatePtr priv = vm->privateData;
    bool running = true;
5452
    virDomainPausedReason reason;
5453
    virSecurityLabelPtr seclabel = NULL;
5454
    virSecurityLabelDefPtr seclabeldef = NULL;
5455
    bool seclabelgen = false;
5456 5457
    virSecurityManagerPtr* sec_managers = NULL;
    const char *model;
5458
    virQEMUDriverConfigPtr cfg = virQEMUDriverGetConfig(driver);
5459
    virCapsPtr caps = NULL;
5460
    bool active = false;
5461
    int ret;
5462 5463 5464 5465

    VIR_DEBUG("Beginning VM attach process");

    if (virDomainObjIsActive(vm)) {
5466 5467
        virReportError(VIR_ERR_OPERATION_INVALID,
                       "%s", _("VM is already active"));
5468
        virObjectUnref(cfg);
5469 5470 5471
        return -1;
    }

5472
    if (!(caps = virQEMUDriverGetCapabilities(driver, false)))
5473
        goto error;
5474

5475 5476 5477 5478 5479
    /* Do this upfront, so any part of the startup process can add
     * runtime state to vm->def that won't be persisted. This let's us
     * report implicit runtime defaults in the XML, like vnc listen/socket
     */
    VIR_DEBUG("Setting current domain def as transient");
5480
    if (virDomainObjSetDefTransient(caps, driver->xmlopt, vm, true) < 0)
5481
        goto error;
5482

5483
    vm->def->id = qemuDriverAllocateID(driver);
5484

5485
    if (virAtomicIntInc(&driver->nactive) == 1 && driver->inhibitCallback)
5486
        driver->inhibitCallback(true, driver->inhibitOpaque);
5487
    active = true;
5488

5489
    if (virFileMakePath(cfg->logDir) < 0) {
5490 5491
        virReportSystemError(errno,
                             _("cannot create log directory %s"),
5492
                             cfg->logDir);
5493
        goto error;
5494 5495 5496
    }

    VIR_FREE(priv->pidfile);
5497
    if (VIR_STRDUP(priv->pidfile, pidfile) < 0)
5498
        goto error;
5499

5500 5501
    vm->pid = pid;

5502
    VIR_DEBUG("Detect security driver config");
5503
    sec_managers = virSecurityManagerGetNested(driver->securityManager);
5504 5505
    if (sec_managers == NULL)
        goto error;
5506 5507

    for (i = 0; sec_managers[i]; i++) {
5508
        seclabelgen = false;
5509 5510
        model = virSecurityManagerGetModel(sec_managers[i]);
        seclabeldef = virDomainDefGetSecurityLabelDef(vm->def, model);
5511
        if (seclabeldef == NULL) {
5512
            if (!(seclabeldef = virSecurityLabelDefNew(model)))
5513 5514 5515
                goto error;
            seclabelgen = true;
        }
5516 5517
        seclabeldef->type = VIR_DOMAIN_SECLABEL_STATIC;
        if (VIR_ALLOC(seclabel) < 0)
5518
            goto error;
5519
        if (virSecurityManagerGetProcessLabel(sec_managers[i],
5520
                                              vm->def, vm->pid, seclabel) < 0)
5521
            goto error;
5522

5523
        if (VIR_STRDUP(seclabeldef->model, model) < 0)
5524
            goto error;
5525

5526
        if (VIR_STRDUP(seclabeldef->label, seclabel->label) < 0)
5527
            goto error;
5528
        VIR_FREE(seclabel);
5529 5530 5531 5532 5533 5534

        if (seclabelgen) {
            if (VIR_APPEND_ELEMENT(vm->def->seclabels, vm->def->nseclabels, seclabeldef) < 0)
                goto error;
            seclabelgen = false;
        }
5535
    }
5536

5537 5538
    if (virSecurityManagerCheckAllLabel(driver->securityManager, vm->def) < 0)
        goto error;
5539 5540 5541
    if (virSecurityManagerGenLabel(driver->securityManager, vm->def) < 0)
        goto error;

5542 5543
    VIR_DEBUG("Creating domain log file");
    if ((logfile = qemuDomainCreateLog(driver, vm, false)) < 0)
5544
        goto error;
5545 5546

    VIR_DEBUG("Determining emulator version");
5547 5548
    virObjectUnref(priv->qemuCaps);
    if (!(priv->qemuCaps = virQEMUCapsCacheLookupCopy(driver->qemuCapsCache,
5549 5550
                                                      vm->def->emulator,
                                                      vm->def->os.machine)))
5551
        goto error;
5552 5553 5554 5555 5556 5557 5558 5559 5560 5561 5562 5563

    VIR_DEBUG("Preparing monitor state");
    priv->monConfig = monConfig;
    monConfig = NULL;
    priv->monJSON = monJSON;

    priv->gotShutdown = false;

    /*
     * Normally PCI addresses are assigned in the virDomainCreate
     * or virDomainDefine methods. We might still need to assign
     * some here to cope with the question of upgrades. Regardless
M
Martin Kletzander 已提交
5564
     * we also need to populate the PCI address set cache for later
5565 5566
     * use in hotplug
     */
5567
    if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE)) {
5568
        VIR_DEBUG("Assigning domain PCI addresses");
5569
        if ((qemuDomainAssignAddresses(vm->def, priv->qemuCaps, vm)) < 0)
5570
            goto error;
5571 5572
    }

5573
    if ((timestamp = virTimeStringNow()) == NULL) {
5574
        goto error;
5575 5576 5577 5578
    } else {
        if (safewrite(logfile, timestamp, strlen(timestamp)) < 0 ||
            safewrite(logfile, ATTACH_POSTFIX, strlen(ATTACH_POSTFIX)) < 0) {
            VIR_WARN("Unable to write timestamp to logfile: %s",
5579
                     virStrerror(errno, ebuf, sizeof(ebuf)));
5580 5581 5582 5583 5584 5585 5586 5587
        }

        VIR_FREE(timestamp);
    }

    qemuDomainObjTaint(driver, vm, VIR_DOMAIN_TAINT_EXTERNAL_LAUNCH, logfile);

    VIR_DEBUG("Waiting for monitor to show up");
5588
    if (qemuProcessWaitForMonitor(driver, vm, QEMU_ASYNC_JOB_NONE, priv->qemuCaps, -1) < 0)
5589
        goto error;
5590

D
Daniel P. Berrange 已提交
5591
    /* Failure to connect to agent shouldn't be fatal */
5592 5593 5594 5595
    if ((ret = qemuConnectAgent(driver, vm)) < 0) {
        if (ret == -2)
            goto error;

D
Daniel P. Berrange 已提交
5596 5597 5598 5599 5600 5601
        VIR_WARN("Cannot connect to QEMU guest agent for %s",
                 vm->def->name);
        virResetLastError();
        priv->agentError = true;
    }

5602
    VIR_DEBUG("Detecting VCPU PIDs");
5603
    if (qemuProcessDetectVcpuPIDs(driver, vm, QEMU_ASYNC_JOB_NONE) < 0)
5604 5605 5606 5607
        goto error;

    VIR_DEBUG("Detecting IOThread PIDs");
    if (qemuProcessDetectIOThreadPIDs(driver, vm, QEMU_ASYNC_JOB_NONE) < 0)
5608
        goto error;
5609 5610 5611

    /* If we have -device, then addresses are assigned explicitly.
     * If not, then we have to detect dynamic ones here */
5612
    if (!virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE)) {
5613
        VIR_DEBUG("Determining domain device PCI addresses");
5614
        if (qemuProcessInitPCIAddresses(driver, vm, QEMU_ASYNC_JOB_NONE) < 0)
5615
            goto error;
5616 5617 5618
    }

    VIR_DEBUG("Getting initial memory amount");
5619
    qemuDomainObjEnterMonitor(driver, vm);
5620 5621 5622 5623 5624 5625 5626
    if (qemuMonitorGetBalloonInfo(priv->mon, &vm->def->mem.cur_balloon) < 0)
        goto exit_monitor;
    if (qemuMonitorGetStatus(priv->mon, &running, &reason) < 0)
        goto exit_monitor;
    if (qemuMonitorGetVirtType(priv->mon, &vm->def->virtType) < 0)
        goto exit_monitor;
    if (qemuDomainObjExitMonitor(driver, vm) < 0)
5627
        goto error;
5628

5629
    if (running) {
5630 5631
        virDomainObjSetState(vm, VIR_DOMAIN_RUNNING,
                             VIR_DOMAIN_RUNNING_UNPAUSED);
5632 5633 5634
        if (vm->def->memballoon &&
            vm->def->memballoon->model == VIR_DOMAIN_MEMBALLOON_MODEL_VIRTIO &&
            vm->def->memballoon->period) {
5635 5636 5637
            qemuDomainObjEnterMonitor(driver, vm);
            qemuMonitorSetMemoryStatsPeriod(priv->mon,
                                            vm->def->memballoon->period);
5638 5639
            if (qemuDomainObjExitMonitor(driver, vm) < 0)
                goto error;
5640 5641
        }
    } else {
5642
        virDomainObjSetState(vm, VIR_DOMAIN_PAUSED, reason);
5643
    }
5644 5645

    VIR_DEBUG("Writing domain status to disk");
5646
    if (virDomainSaveStatus(driver->xmlopt, cfg->stateDir, vm) < 0)
5647
        goto error;
5648

5649 5650
    /* Run an hook to allow admins to do some magic */
    if (virHookPresent(VIR_HOOK_DRIVER_QEMU)) {
5651
        char *xml = qemuDomainDefFormatXML(driver, vm->def, 0);
5652 5653 5654 5655 5656 5657 5658 5659 5660 5661 5662
        int hookret;

        hookret = virHookCall(VIR_HOOK_DRIVER_QEMU, vm->def->name,
                              VIR_HOOK_QEMU_OP_ATTACH, VIR_HOOK_SUBOP_BEGIN,
                              NULL, xml, NULL);
        VIR_FREE(xml);

        /*
         * If the script raised an error abort the launch
         */
        if (hookret < 0)
5663
            goto error;
5664 5665
    }

5666 5667
    VIR_FORCE_CLOSE(logfile);
    VIR_FREE(seclabel);
5668
    VIR_FREE(sec_managers);
5669
    virObjectUnref(cfg);
5670
    virObjectUnref(caps);
5671 5672 5673

    return 0;

5674 5675
 exit_monitor:
    ignore_value(qemuDomainObjExitMonitor(driver, vm));
5676
 error:
5677 5678 5679 5680 5681 5682
    /* We jump here if we failed to attach to the VM for any reason.
     * Leave the domain running, but pretend we never attempted to
     * attach to it.  */
    if (active && virAtomicIntDecAndTest(&driver->nactive) &&
        driver->inhibitCallback)
        driver->inhibitCallback(false, driver->inhibitOpaque);
5683 5684
    VIR_FORCE_CLOSE(logfile);
    VIR_FREE(seclabel);
5685
    VIR_FREE(sec_managers);
5686 5687
    if (seclabelgen)
        virSecurityLabelDefFree(seclabeldef);
5688
    virDomainChrSourceDefFree(monConfig);
5689
    virObjectUnref(cfg);
5690
    virObjectUnref(caps);
5691 5692 5693 5694
    return -1;
}


5695
static virDomainObjPtr
5696 5697 5698
qemuProcessAutoDestroy(virDomainObjPtr dom,
                       virConnectPtr conn,
                       void *opaque)
5699
{
5700
    virQEMUDriverPtr driver = opaque;
5701
    qemuDomainObjPrivatePtr priv = dom->privateData;
5702
    virObjectEventPtr event = NULL;
5703
    unsigned int stopFlags = 0;
5704

5705
    VIR_DEBUG("vm=%s, conn=%p", dom->def->name, conn);
5706

5707 5708 5709
    if (priv->job.asyncJob == QEMU_ASYNC_JOB_MIGRATION_IN)
        stopFlags |= VIR_QEMU_PROCESS_STOP_MIGRATED;

5710 5711
    if (priv->job.asyncJob) {
        VIR_DEBUG("vm=%s has long-term job active, cancelling",
5712
                  dom->def->name);
5713
        qemuDomainObjDiscardAsyncJob(driver, dom);
5714 5715
    }

5716 5717
    if (qemuDomainObjBeginJob(driver, dom,
                              QEMU_JOB_DESTROY) < 0)
5718 5719 5720
        goto cleanup;

    VIR_DEBUG("Killing domain");
5721

5722
    qemuProcessStop(driver, dom, VIR_DOMAIN_SHUTOFF_DESTROYED, stopFlags);
5723

5724
    virDomainAuditStop(dom, "destroyed");
5725
    event = virDomainEventLifecycleNewFromObj(dom,
5726 5727
                                     VIR_DOMAIN_EVENT_STOPPED,
                                     VIR_DOMAIN_EVENT_STOPPED_DESTROYED);
5728

5729 5730
    qemuDomainObjEndJob(driver, dom);

5731
    if (!dom->persistent) {
5732
        qemuDomainRemoveInactive(driver, dom);
5733 5734
        dom = NULL;
    }
5735

5736
    if (event)
5737
        qemuDomainEventQueue(driver, event);
5738

5739
 cleanup:
5740
    return dom;
5741 5742
}

5743
int qemuProcessAutoDestroyAdd(virQEMUDriverPtr driver,
5744 5745 5746
                              virDomainObjPtr vm,
                              virConnectPtr conn)
{
5747
    VIR_DEBUG("vm=%s, conn=%p", vm->def->name, conn);
5748 5749
    return virCloseCallbacksSet(driver->closeCallbacks, vm, conn,
                                qemuProcessAutoDestroy);
5750 5751
}

5752
int qemuProcessAutoDestroyRemove(virQEMUDriverPtr driver,
5753 5754
                                 virDomainObjPtr vm)
{
5755
    int ret;
5756
    VIR_DEBUG("vm=%s", vm->def->name);
5757 5758 5759
    ret = virCloseCallbacksUnset(driver->closeCallbacks, vm,
                                 qemuProcessAutoDestroy);
    return ret;
5760
}
5761

5762
bool qemuProcessAutoDestroyActive(virQEMUDriverPtr driver,
5763 5764
                                  virDomainObjPtr vm)
{
5765
    virCloseCallback cb;
5766
    VIR_DEBUG("vm=%s", vm->def->name);
5767
    cb = virCloseCallbacksGet(driver->closeCallbacks, vm, NULL);
5768
    return cb == qemuProcessAutoDestroy;
5769
}